DE102008023906B4 - Method and system for the automatic generation of throughput models for semiconductor systems - Google Patents

Abstract

Throughput model generator (100) comprising: a data input section (101) adapted to receive plant-specific information about configuration and operational status of a process plant (110, 110A, 110B, 110C) used in a manufacturing environment (150), the process plant ( 110, 110A, 110B, 110C) comprises a plurality of functional units (111,.., 115) and wherein the plant-specific information comprises plant events which are generated by the plurality of functional units (111,.., 115) of the process plant (110, 110A, 110B, 110C ) are generated;a database (104) operatively connected to the data input section (101) and configured to store a data model of the process plant (110, 110A, 110B, 110C), the data model corresponding to at least one material flow in the process plant (110 , 110A, 110B, 110C); a model calculator (102) operatively connected to the database (104) and configured to represent a throughput of the process plant (110, 110A, 110B , 110C) based on the data model and one or more throughput-related parameters;a user interface (106) operatively connected to the model calculator (102) and configured to select a value of the one or more throughput-related parameters;a data pre-processing module ( 103), which is functionally connected to the data input section (101) and the database (104) and designed to determine process steps and associated process times for the material flow in the process plant (110, 110A, 110B, 110C), the data pre-processing module (103 ) is further configured to determine the process time of each process step by detecting a plurality of plant events for each process step contained in the plant-specific information and by determining a representative time interval for each process step, and wherein the data pre-processing module (103) is further configured to determine a first representative ive time interval for a first process step not related to a process recipe and a second representative time interval for a second process step related to a process recipe; and wherein the model calculator (102) is further adapted to determine the determine throughput of the process plant (110, 110A, 110B, 110C) based on the data model with respect to the first and second representative time intervals.

Description

Field of the present invention

In general, the present disclosure relates to the field of the production of microstructures, for example of integrated circuits, and in particular relates to the throughput properties of complex process plants, for example cluster or multi-process plants, which are used for the production of semiconductor components or other microstructures.

Description of the prior art

Today's global market forces manufacturers of mass products to offer them at high quality and low price. It is therefore important to improve the yield and the process efficiency in order to minimize the manufacturing costs. This applies in particular in the field of microstructure production, for example for the production of semiconductor components, since in this area it is important to combine the most modern technology with mass production processes. It is therefore the goal of semiconductor manufacturers, or of manufacturers of microstructures in general, to reduce the consumption of raw materials and consumables, while at the same time improving the utilization of the process equipment. The last-mentioned aspect is particularly important since modern semiconductor manufacturing facilities require systems that are extremely costly and represent the major part of the total manufacturing costs. At the same time, the process equipment in the semiconductor manufacturing facility has to be replaced more frequently than in most other technical fields due to the rapid development of new products and processes, which therefore make appropriately adapted process equipment necessary.

DE 10 2006 004 413 A1 discloses manufacturing processes, in particular the scheduling or scheduling of product flows in a manufacturing environment, such as a semiconductor manufacturing facility, in which several different product types and process and measuring systems are handled.

DE 10 2004 036 300 A1 discloses process control systems and in particular the use of profitability calculations for easier or better control of a process or a plant.

DE 11 2005 002 474 T5 discloses an industrial process and in particular methods and systems for adjusting the measurement data acquisition based on the available measurement capacity.

DE10 2006 004 408 A1 discloses the monitoring of systems that are required for the processing of different types of semiconductor components with different process recipes.

LINDEMANN, M .; SCHMID, S .: Market overview: Simulation tools in production and logistics. In: PPS Management, 2007,12,28-35, discloses a market study with regard to various software solutions for the creation of simulation models in production and logistics.

US 6,873,878 B2 discloses a throughput monitoring and analysis system and method comprising: factor retrieval means and terminal means, the factor retrieval means retrieving a plurality of raw time data of the production factors corresponding to selected operational events during the execution of operational events of the complex machine, the terminal stores the time data and displays the related information of the factors of production and the monitoring results according to the required condition set and the data and information.

Integrated circuits are typically produced in automated or semi-automated production facilities, with a large number of process and measurement steps going through until the component is completed. The number and type of process steps and measurement steps that a semiconductor component has to go through depends on the circumstances of the semiconductor component to be manufactured. A normal process flow for an integrated circuit contains a large number of photolithography steps in order to transfer a circuit pattern for a special component level into a lacquer layer, which is subsequently structured in order to form a lacquer mask for further processes in the structuring of the component layer under consideration, for example by etching processes or Implantation processes and the like are carried out. Thus, layer after layer, a large number of process steps are carried out on the basis of a special lithography mask set for the various layers of the specified component. For example, a modern CPU requires several hundred process steps, each of which must be carried out within specified process limits in order to meet the specifications for the component under consideration. Since the majority of the process areas are component-specific, many of the measuring processes and the actual manufacturing processes are specially designed for the component under consideration and require special parameter settings on the respective measuring and process systems.

A variety of different Product types processed simultaneously, such as memory chips with different designs and storage capacities, CPUs with different designs and operating speeds, and the like, with the number of different product types 100 or more can be achieved in manufacturing facilities for the manufacture of application-specific IC's (ASIC). Since each of the different product types requires a special process sequence, possibly based on a different mask set for lithography, special settings are required in the various process systems, such as deposition systems, etching systems, implantation systems, CMP (chemical-mechanical polishing) systems, and the like. This means that a large number of different system parameter settings and product types occur simultaneously in a production environment.

In the following, the parameter setting for a special process in a specified process plant or measuring or inspection plant is generally referred to as a process recipe or simply a recipe. Thus, a large number of different process recipes are required even for the same type of process plant, which are to be applied to the process plants at the point in time in which the corresponding product types are processed in the respective plants. However, the sequence of process recipes that are executed in the process and measuring systems or in functionally combined system groups, as well as the recipes themselves, are subject to frequent changes due to rapid product changes and very variable processes involved. As a result, the system behavior is a very critical production parameter, particularly with regard to throughput, as this significantly influences the overall production costs of the individual components. The course of the throughput over time for individual processes and measuring systems or even for certain units, such as process modules, substrate robots, charging stations and the like, can, however, due to the complexity of the production sequences, which contain a large number of product types and a correspondingly large number of processes, which in turn are subject to frequent recipe changes and remain unobserved for a long time.

In the recent past, process systems have become increasingly complex in that a process system can contain several functional modules or units, these systems being referred to as cluster or cluster systems that work in parallel and / or sequentially, so that products arriving at the cluster system are therein corresponding to a large number of processes or plant-internal process sequences are processed depending on the process recipe and the current plant status. The cluster system enables a sequence of correlated processes to be carried out, the overall efficiency being increased by, for example, reducing transport activities within the production facility and / or by increasing the system capacity and availability by using several parallel process chambers for the same process step. Due to the large number of units involved, the mutual dependence of process recipes, system-internal process paths and other parameters, such as the delivery strategy to the system using the global transport system, the configuration of the interface between the global transport system and the system, i.e. H. the number of charging stations and the like, very difficult to evaluate, for example, with regard to the overall system performance in terms of system throughput.

Consequently, the overall system behavior, for example with regard to yield and throughput, is a very critical and complex production parameter, since this significantly influences the overall production costs of the individual components. The course of the system yield and the throughput over time for individual process and measuring systems or even certain units thereof, such as process modules, substrate robots, charging stations and the like, can, due to the complexity of the production sequences with the large number of product types and a correspondingly large number of processes which in turn are subject to frequent recipe changes, can remain unobserved for a long time. This means that systems with poor performance may go undetected for a long time if the performance of a system group to which the system in question belongs lies within its usual performance limit, which is typically selected so that a relatively wide range of fluctuations due to the complexity of the Processes and the systems involved is possible.

For this reason, it is of great importance for the semiconductor manufacturer to observe and determine corresponding measures that provide a measure of the performance of individual process systems, which also makes it possible for system suppliers to use software and hardware components of process systems specifically on the basis of the Improve data provided by manufacturers. Since the plant requirements depend heavily on the manufacturer-specific conditions, a large number of industry standards have been defined in order to create a basis for defining a common global set of semiconductor plant requirements, whereby company-specific requirements for production plants are reduced, while on the side of the plant suppliers the energy is focused on improving the Process idiosyncrasies can instead be concentrated on maintaining many internal products. Thus, in some industrial fields, there have been a variety Plant-specific standards that relate to the definition of plant messages known to the semiconductor industry and the SECS (Semi (Semiconductor Plant and Material Institute) Plant Communication Standard), which provides a common language for communication between process plants and remote control systems. Similarly, a variety of standards have been established to define plant behavior. For example, in the field of semiconductors, the E10 and E58 standards provide a basis for evaluating the reliability, availability and maintenance properties (RAM) of process plants using standard plant states. Other standards, such as the E116 standard, were introduced to describe the performance behavior of process plants on the basis of a state model, whereby the plant state is automatically exposed by providing state transitions and runtime information.

Consequently, in a complex production environment, a large amount of information relating to the process equipment is available with a different degree of resolution depending on the overall configuration of a control system of the production environment under consideration. In other words, the corresponding state models for representing individual process systems in the production environment enable the process systems to be monitored and controlled at a higher abstract level, for example by means of a higher-level MES (production execution system), while at a lower level information itself with regard to the operating status of individual units of the various process systems are available in the form of the corresponding system messages. In this regard, a functional unit of a process plant is to be understood as a unit that can hold a workpiece, for example a substrate for the production of semiconductor components, for example a process chamber, a robot, and the like. Although a large amount of valuable system-specific information is available in the production environment, an assessment and monitoring or control of throughput-related aspects of the process systems is difficult, since unobserved throughput losses typically occur in the process lines, for example due to process and setting changes and / or system malfunctions and the like be evoked. In other words, due to the complexity of modern semiconductor manufacturing facilities, which contain a certain degree of variability in the manufacturing processes, average throughput rates, which are calculated from the batch process side or from status output data, do not allow the system performance behavior of individual systems or system groups with regard to throughput to be efficiently recognized . For example, in many process plants a large number of the functional units are process chambers for carrying out a corresponding process on the basis of an assigned process recipe. In many process systems, two or more equivalent process chambers are provided in order to achieve a certain degree of parallelism when the substrates are processed according to the specified recipe. In a similar way, different process steps can be integrated into a single system, which is then also referred to as a cluster system, so that a large number of process steps are carried out on the basis of different process recipes, while other functional units are responsible for providing the required substrate handling resources, for example in Form of charging stations, robots, and the like. Thus, the throughput of a complex process plant can essentially depend on the basic structure with regard to the substrate handling resources and the actual process modules in combination with the special process recipes that are to be used in the various process modules.

For example, if one of several parallel or equivalent process modules fails or if one of several parallel or equivalent process modules fails completely, the process system continues to function, but has a lower throughput, which may be determined by determining the throughput of system groups, for example on a monthly basis, as is often the case in semiconductor factories may go undetected. Even if the system messages are monitored in order to identify the current functional status of the process system in question, however, an estimate of the throughput is made on the basis of a series of lots or substrates up to a specific point in time and the corresponding time required to process the substrates or lots required is achieved. Thus, only a “retrospective” estimate of the throughput is obtained, while the “current” throughput is difficult to estimate, especially if the functional status of the process plant under consideration has changed. In other words, the throughput can depend on the current status of the process plant and the respective process that is to be used in the various functional units, so that conventional procedures that are based on the number of substrates that have passed the past Time interval have been processed and are dependent on the length of the time period, do not allow a predictive estimate of the throughput, and these solutions also do not allow an estimate of the throughput behavior for one or more process plants with regard to different operating scenarios in order to evaluate the throughput when certain events occur , for example a change in process recipes, one Reorganization of the operational status of the process plant, and the like.

For this purpose, throughput models have been proposed in order to achieve a certain degree of “predictive” function with regard to the throughput of complex cluster systems. A throughput model is to be understood as a mathematical method or a model in order to determine the throughput of special process plants when these are operated on the basis of specified process recipes, whereby a number of conditions or parameters are also taken into account. The throughput is expressed by the number of machined workpieces divided by the time interval required to machine these workpieces. In these models it is further assumed that before and after this time interval the process plant is empty, i. H. the corresponding units of the process systems, such as the transport robots, process chambers, and the like, do not contain any workpieces. The number of workpieces processed within the period of time under consideration is often referred to as a cascade, with the individual elements of a cascade representing individual workpieces, e.g. Group of substrates takes place. In this case, the corresponding process systems are also referred to as stack systems.

At the beginning of a cascade, i. H. When the first workpiece is fed to the processing system, a certain time elapses before the first workpiece is completely machined. This time, d. H. the time it takes for the first workpiece to go through all the process steps that are to be performed for the specified workpiece is referred to as the "time for the first workpiece", and this time interval is typically longer than the respective time intervals with which the subsequent ones Workpieces leave the last process step in the process plant. Thus, in a corresponding throughput model, the throughput is determined by the length of a cascade divided by the total process time, which is the time for the first piece, i.e. H. contains the process time for the first workpiece in the cascade, and the intervals for the remaining workpieces in the cascade. Thus, in a throughput model, the conditions and the parameters that determine the time for the first piece and the time intervals for the subsequent pieces must be differentiated and quantified. This can be done in two different ways. For the time for the first piece, the actual process sequence in the process plant under consideration is determined, i. H. the actual sequence of processes that are to be carried out in the individual process chambers on the basis of corresponding process recipes. On the other hand, the system resources are the decisive point for the time intervals of the subsequent workpieces. In this case, it is typically necessary to consider all resources in order to determine a limiting or a bottleneck resource of the process plant that determine the time intervals for the subsequent workpieces.

Thus, the determination of anticipated throughput models can require a certain amount of intuition and a high level of experience on the part of the process engineer, in that a corresponding very limited information is filtered out of a limited number of system messages and the material flows in the process system under consideration are suitably defined. Thus, the conventional solution for generating a throughput model is a very time-consuming task, the result of which depends on very subjective criteria, which can lead to inconsistencies when using throughput models that have been created by different process engineers. Furthermore, the time required to generate the model leads to long time intervals between updates of the throughput models, which can lead to the use of possibly outdated models.

In view of the situation described above, the present disclosure relates to systems and methods for determining a throughput of process plants in a complex manufacturing environment, wherein one or more of the problems identified above are avoided or at least reduced.

Overview of the revelation

In general, the subject matter disclosed herein relates to systems and methods for estimating the throughput of process equipment in a manufacturing environment, which in some illustrative embodiments represents a manufacturing environment for manufacturing microstructure devices in which equipment-level equipment messages are used to obtain statistically relevant equipment-specific data that can be used to evaluate the throughput of the process plant for a specified functional scenario. On the basis of the system messages with higher resolution, i.e. on the basis of system messages that represent system events that were generated by each functional unit involved in a process system under consideration, corresponding material flows in the process system and the associated process time intervals can be determined in an automated manner on the basis of high statistical relevance be obtained, thus obtaining the corresponding Throughput estimates are given a high degree of robustness. In some illustrative aspects disclosed herein, throughput models are generated automatically by determining the functional units, such as the robot transport units, process chambers, and the like, on the basis of plant messages that are obtained with high coverage of the respective process plants, ie the plant messages are obtained from the plant interface, whereby an improved resolution compared to, for example, a representation of the respective process systems in a higher-level control system, such as an MES, as it is typically used in complex semiconductor manufacturing facilities, in which the functional units are actually not "instantiated" in the corresponding state models that represent the systems " are. Furthermore, the material processes can be determined by the cluster system or by a collection of the systems under consideration at the functional unit level, with an automatic distinction being made between functional units that are not linked to a process recipe and functional units, such as process chambers, that are linked to a process recipe . Consequently, in some illustrative solutions, mean process time intervals are determined for recipe-dependent units, which can be done on the basis of suitable determined time intervals that are determined from system events, while statistically relevant process time intervals are also determined for functional units that are not related to process recipes. From these time intervals with high statistical significance and on the basis of the previously determined material processes, a suitable throughput model is automatically created with a high degree of objectivity, which is then used to estimate throughput values with regard to preselected operating scenarios or to estimate the current throughput status of the systems in the Manufacturing environment can be used.

One illustrative system disclosed herein relates to a throughput model generator. The throughput model generator having: a data input section which is designed to receive plant-specific information about the configuration and operating status of a process plant that is used in a production environment, wherein the process plant comprises a plurality of functional units and wherein the plant-specific information comprises plant events that are generated by the plurality of functional units of the process plant be generated; a database which is functionally connected to the data input section and is designed to store a data model of the process plant, the data model representing at least one material flow in the process plant; a model calculator which is functionally connected to the database and is designed to determine a throughput of the process plant on the basis of the data model and one or more throughput-related parameters; a user interface operatively connected to the model calculator and configured to select a value of the one or more throughput-related parameters; a data preprocessing unit which is functionally connected to the data input section and the database and is designed to determine process steps and associated process times for the material flow in the process system, the data preprocessing unit also being designed to determine the process time of each process step by recognizing multiple system events for each process step that occurs in the system-specific information is contained, and by determining a representative time interval for each process step, and wherein the data preprocessing unit is further designed, a first representative time interval for a first process step that is not related to a process recipe, and a second representative time interval for determine a second process step related to a process recipe; and wherein the model calculator is further configured to determine the throughput of the process plant on the basis of the data model in relation to the first and the second representative time interval.

One illustrative method disclosed herein relates to the automatic determination of a throughput of a process plant. The method for automatically determining a throughput of a process plant comprises: recognizing all functional units of the process plant by monitoring plant messages which are generated by the process plant during operation; Recognizing one or more material flows as a sequence of process steps that are carried out by at least some of the functional units; Determining a process time interval for each of the process steps of the one of the plurality of material flows; and determining a throughput of the process plant for a selected process scenario for the one or more material flows using the determined process time intervals, wherein determining a process time interval for each of the process steps comprises: recognizing first functional units that are not dependent on an assigned process recipe, and recognizing Functional units that are dependent on an assigned process recipe, the first and second functional units being recognized using at least the system messages; and wherein the process interval is determined on the basis of the identified first functional units and second functional units.

Figure list

• 1a schematisch eine Fertigungsumgebung mit einer anschaulichen Cluster-Anlage zeigt, deren Durchsatz auf der Grundlage eines automatisch erzeugten Durchsatzmodells zu bestimmen ist;
• 1b schematisch ein Zeitdiagramm zur Bearbeitung mehrerer Substrate, die auch als eine Kaskade bezeichnet werden, in der Cluster-Anlage aus 1a zeigt;
• 1c schematisch ein Diagramm eines Materialablaufs durch die Cluster-Anlage aus 1a zeigt;
• 1d schematisch eine Fertigungsumgebung mit mehreren Prozessanlagen zeigt, die funktionsmäßig mit einem automatischen Durchsatzmodellgenerator gemäß anschaulicher Ausführungsformen verbunden sind;
• 1e schematisch eine Fertigungsumgebung mit einem automatischen Durchsatzmodellgenerator gemäß noch weiterer anschaulicher Ausführungsformen zeigt;
• 1f schematisch ein Datenmodell mit vorverarbeiteten durchsatzbezogenen Daten zeigt, wie sie durch einen automatischen Durchsatzmodellgenerator gemäß anschaulicher Ausführungsformen verwendet werden;
• 1g schematisch einen Entscheidungsbaum zeigt, wie er in dem automatischen Durchsatzmodellgenerator zum Erkennen eines Materialsflusses in der Prozessanlage gemäß weiterer anschaulicher Ausführungsformen eingerichtet ist;
• 1h schematisch ein Zeitdiagramm von Anlagennachrichten und entsprechenden Anlagenereignissen zeigt, die zum Bestimmen von Prozesszeitintervallen gemäß anschaulicher Ausführungsformen angewendet werden;
• 1i und 1j schematisch Zeitdiagramme miteinander in Beziehung stehender Funktionseinheiten in einer Prozessanlage zeigen;
• 1k schematisch ein Diagramm darstellt, das die Häufigkeit von Prozesszeitintervallen für eine spezifizierte Funktionseinheit repräsentiert, um ein statistisch signifikantes repräsentatives Prozesszeitintervall für die betrachtete Funktionseinheit gemäß anschaulicher Ausführungsformen zu erhalten;
• 1l schematisch eine Zustandsmaschine zeigt, um die Gültigkeit eines Ablaufs von Anlagenereignissen gemäß weiterer anschaulicher Ausführungsformen zu verifizieren; und
• 1m schematisch Datenpunkte zeigt, wie sie in dem Datenmodell verwendet werden, um eine Geschichte von durchsatzbezogenen Informationen gemäß noch weiterer anschaulicher Ausführungsformen einzurichten.

Further embodiments of the present disclosure are defined in the appended claims and will become more clearly apparent from the following detailed description when studied with reference to the accompanying drawings, in which:

• 1a shows schematically a production environment with an illustrative cluster system, the throughput of which is to be determined on the basis of an automatically generated throughput model;
• 1b schematically shows a time diagram for processing several substrates, which are also referred to as a cascade, in the cluster system 1a indicates;
• 1c schematically shows a diagram of a material flow through the cluster system 1a indicates;
• 1d schematically shows a manufacturing environment with multiple process plants operatively connected to an automatic throughput model generator according to illustrative embodiments;
• 1e schematically shows a manufacturing environment with an automatic throughput model generator according to still further illustrative embodiments;
• 1f schematically shows a data model with preprocessed throughput related data as used by an automatic throughput model generator according to illustrative embodiments;
• 1g schematically shows a decision tree as it is set up in the automatic throughput model generator for recognizing a material flow in the process plant according to further illustrative embodiments;
• 1h schematically depicts a timing diagram of facility messages and corresponding facility events used to determine process time intervals in accordance with illustrative embodiments;
• 1i and 1y schematically show timing diagrams of interrelated functional units in a process plant;
• 1k schematically shows a diagram which represents the frequency of process time intervals for a specified functional unit in order to obtain a statistically significant representative process time interval for the functional unit under consideration according to illustrative embodiments;
• 1l schematically shows a state machine for verifying the validity of a sequence of system events according to further illustrative embodiments; and
• 1m schematically shows data points as they are used in the data model to establish a history of throughput-related information according to still further illustrative embodiments.

Detailed description

Although the present disclosure has been described with reference to the embodiments as illustrated in the following detailed description and drawings, it should be noted that the following detailed description and drawings are not intended to provide the present disclosure to the specific illustrative embodiments disclosed to restrict, but the descriptive embodiments are only exemplary represent the appended claims.

In general, the present disclosure relates to systems and methods for estimating the throughput of one or more process plants in a manufacturing environment based on a throughput model that is established in an automated manner by using an appropriate amount of plant-specific information generated by the process plants during operation in order to obtain statistically relevant process time intervals for individual process steps that are carried out in the process plants under consideration. Furthermore, efficient algorithms are set up in order to automatically determine the functional units of a process plant and corresponding material processes or material flows using at least some of the recognized functional units. As a result, the functional status of a process system under consideration can be determined automatically at the functional unit level, with the material processes or flows therein also being recognized in order to enable the determination of recipe-dependent process steps and recipe-independent process steps for the system under consideration. As a result, recipe-specific properties of a specified operating scenario can be taken into account in an automated manner, with the system-dependent resources also being represented by the functional units that do not depend on a specific recipe in order to enable resource-dependent process time intervals to be determined. In this way, the influence of recipe changes, of information about the system configuration, for example with regard to the addition of additional resources, the switching off of units due to maintenance and the like can be estimated with improved accuracy and robustness with regard to process fluctuations on the basis of the statistically determined process time intervals. Compared to conventional strategies, the generation of corresponding throughput models is carried out with a high degree of consistency, since corresponding criteria for determining throughput-specific parameters are applied in the same way to a large number of process plants, whereby the possibility is also created to vary the criteria in order to thus obtaining different "views" with regard to the generation of throughput models. For example, the statistical data processing can be changed by privileged users, for example, in order to recognize a dependency of the throughput results on data processing processes and / or to adapt or "calibrate" the estimated throughput models to actual measurement data, and the like in order to improve the accuracy of the throughput estimates for the to improve the current situation in the manufacturing environment. For example, the process time intervals for static functional units, ie units that are not dependent on special process recipes, can be varied in order to improve the agreement with actually measured throughput values and / or to estimate a throughput improvement by redesigning the corresponding functional units, for example by increasing the transport capacity is increased and the like. Thus, on the basis of the automatically generated throughput models, which are based on statistically relevant data, a large number of operating scenarios can be examined, with corresponding results being able to be used to configure and / or control at least part of the production environment under consideration. For example, the maintenance of functional units of process systems can be coordinated in order to obtain a minimized loss of throughput by appropriately planning the total flow of material in the production environment on the basis of the throughput estimates. That is to say, when examining the throughput dependency for a large number of process systems for several product types, recipes and the like, a process scenario can be determined which has a minimal throughput loss when maintaining one or more of the functional units. Similarly, if a functional unit malfunctions, the impact on overall throughput of a portion of the manufacturing environment can be determined with little or substantially no delay, and if necessary, the entire flow of materials in the manufacturing environment can be re-scheduled to increase overall throughput at a high level.

Furthermore, in some illustrative aspects disclosed herein, the throughput model generator is based on historical data that enable the detection of functional states of process plants that may not be recognizable by currently available plant messages, whereby the responsiveness of the throughput models with regard to changes in the plant status is significantly improved. For example, process plant messages that have been collected over a certain operating period can identify a special functional status of the process plant, for example a functional unit is switched off due to maintenance and the like, which is therefore not reflected in the throughput model of the process plant currently in use. When the functional unit is functional again, the change in the status of the system is recognized and a suitable material flow, which now again contains the newly connected functional unit, is called up on the historical data in combination with the corresponding historical process time intervals. When a new configuration of the process plant under consideration occurs, the throughput model can consequently be adapted in a corresponding manner and thus the throughput for the current process scenario or another assumed process scenario can be determined with a high degree of accuracy.

Thus, according to the principles disclosed herein, the physical conditions, i. H. the number and arrangement of functional units, the type of workpieces to be processed, such as individual substrates, groups of substrates and the like, the process time intervals defined by the process recipes, d. H. the duration of the actual process steps, the process time intervals that are determined by mechanical requirements, such as robot movement times, different pen movement times for receiving a disc in a process chamber and the like, and the flow of substrates through the process system, d. H. the steps and their relationship to the functional units that execute the steps are determined in an automated manner on the basis of suitable control sequences and statistically relevant system-specific data. Thus, other throughput-related parameters for defining various operating scenarios can be fed into the systems in the method according to the process requirements and the like. For example, the length of a special cascade, the lot size and the like can be used as throughput-related parameters. Throughput values can thus be determined with a high degree of accuracy on the basis of robust statistical data, as a result of which throughput properties can be determined during a desired operating period and for a process scenario.

With reference to the accompanying drawings, further illustrative embodiments will now be described in more detail.

1a shows schematically a manufacturing environment 150 , which, in one illustrative embodiment, represents a manufacturing facility for the production of microstructure components, such as modern semiconductor components and the like. The environment 150 includes one or more process plants 110 that are designed to carry out a sequence of process steps that are required for machining workpieces that can represent individual workpieces or a group of workpieces, depending on the overall configuration of the system 110 . In the following, substrates are referred to as workpieces, which represent wafers and the like, which are designed in a suitable manner in order to produce microstructure components therein and thereon. It should be noted, however, that the principles disclosed herein apply to any type of complex product that requires a variety of process steps. It should be noted, therefore, that the present disclosure should not be viewed as limited to microstructure devices and corresponding substrates unless such limitations are specifically set forth in the appended claims.

The manufacturing environment 150 includes a transport system 120 , which is provided in the form of an automated system in which vehicles, workpieces or substrates 130 from one process plant to another depending on the entire process in the area 150 transport. The transport system also includes 120 suitable interface systems for exchanging the substrates 130 , which are also contained in suitable containers, with one or more processing systems 110 . To this end, the one or more process tools include 110 one or more charging stations 111 that are formed, the substrates 130 from the transport system 120 recording and also removing substrates that are in the plant 110 edited to enable. Furthermore, in the example shown, represents the process plant 110 a cluster system, ie a process system that has several process chambers 112 , 113 , 114 contains, some of which are designed to carry out the same process, while other process chambers are assigned to different processes. If, for example, several etching steps are to be carried out with different etching chemistries, at least some of the process chambers form 112 , 113 and 114 a corresponding sequence of different etching steps, each of which is carried out on the basis of a different process recipe. In other cases, the system will function in parallel and sequentially 110 provided, for example when two or more process chambers perform essentially the same process, while at least one further process chamber performs a different process. As an example, assume that the process plant 110 a sequence of process steps in which a substrate executes the chamber 112 and one of the chambers 113 or 114 because these process chambers are operated on the basis of the same process recipe.

Furthermore, the process plant includes 110 an internal transport system 115 , for example in the form of a robot system, which is designed, substrates to the respective units 111 , 112 , 113 and 114 to be transported and removed from there. It should be noted that in the context of the present disclosure, the process chambers 112 , 113 , 114 , the charging station 111 and the internal transport system 115 can be referred to as a functional unit, which represents a base unit with the ability to hold a substrate. Furthermore, the functional units 111 , ..., 115 operated under the control of an in-plant control system (not shown), with corresponding activities of the units 111 , ..., 115 , which are also referred to as plant events, are transmitted to the plant-internal control system in response to corresponding commands issued by the control system. This means that corresponding system messages, such as the activity of the individual functional units, can be accessed 111 , ..., 115 can be accessed by an external device via a corresponding system interface, as described in more detail below.

The process plant receives during operation 110 a cascade of substrates 120 , ie appropriate substrates are in the plant 110 fed, ie the one or more charging stations 111 , by means of the transport system 120 so that a predefined number of substrates 120 continuously in the system 110 can be edited. For this purpose, one or more containers with a corresponding number of substrates 130 the charging station 111 supplied, or corresponding substrates 130 or corresponding groups on substrates are essentially sequentially assigned to the charging station 111 so fed that a continuous supply of the process chamber 112 showing the first process of material flow in the facility 110 represented, is guaranteed. Thus, the capacity of the charging station (s) 111 , for example number of individual charging stations and the like, the transport properties of the system 120 adapted so that cascading operation in the system 110 is achieved.

1b shows schematically the material flow in the process plant 110 according to the example specified above, in which a first process in the chamber 112 based on a first process recipe and a second process in the chamber 113 or 114 is carried out on the basis of a second process recipe. Thus, the material flow 140 seven steps 141 , ..., 147 have, each of which has at least one functional unit 111 , ..., 115 assigned. For example, represents the first step 141 of the river 140 a substrate that is on the charging station 111 is arranged in which one of the substrates 130 for inclusion by the transport system 115 is being prepared. Thus, a special process time interval can be used with the step 141 depending on the structure of the charging station 111 be linked. The second step 142 represents the movement of the transport system 115 for receiving a substrate and for transporting the substrate to the process chamber 112 . In the third step 143 becomes the substrate in the chamber 112 processed on the basis of a special process recipe. So the step depends 143 of the structure of the chamber 112 and the process recipe under consideration. However, it should be noted that the step 143 may have corresponding “partial steps”, which can relate to steps that are essentially independent of the recipe, such as receiving the substrate in the chamber without the substrate actually being processed and preparing the substrate so that it can be used by the transport system 115 is recorded after the actual process has ended. After that, the step concerns 144 an activity of the plant-internal transport system 115 to put the substrate in the chamber 112 and take this to one of the chambers 113 and 114 to transport. Thus, the fifth step 145 represent a "parallel" step, as both chambers are operated on the basis of the same process recipe. Next concerns the step 146 in turn, a transport activity of the system 115 to put the substrate in one of the chambers 113 , 114 pick up and the substrate of the charging station 111 to feed. Thus represents the last step 147 a corresponding "activity" of the charging station 111 in order, for example, to pick up the substrate and prepare it for the arrival of another substrate.

During the operation of the process plant 110 for a given cascade, ie a given number of substrates 130 , is the throughput of the plant 110 for this cascade is determined by the number of charge-containing substrates and the time required to pass through the substrate in accordance with the material flow 140 is required. As explained above, a corresponding cascade is divided into different time periods in order to enable a distinction to be made between process-specific or recipe-specific aspects and resource-specific aspects.

1c shows schematically a time diagram for processing a cascade from the substrates, wherein a time period FPT represents the time for the first workpiece, which corresponds to the total process time of the first substrate, while the time period, which are designated as “PI”, the leakage intervals of the substrates after the last step 147 correspond, ie after arrival at the charging station 111 . The time interval for processing the cascade is thus defined by the time for the first piece FPT plus 9 close to the workpiece interval PI. Hence the throughput of the plant 110 given by 10 divided by the total processing time, as previously explained. According to the principles disclosed herein, material flows, such as the flow 140 and the process time intervals associated with each step and sub-step of the material flow are determined on the basis of system messages in an automated manner by using statistical data processing in order to obtain robust and consistent values for the respective process times. It should be noted that several different material flows can be created within a special process system, which can depend on the current system configuration and / or the overall process strategy for processing the substrates under consideration. For example, as previously explained, there are typically a large number of different product types in the manufacturing environment 150 available, whereby the various product types can differ with regard to the process strategy. In this case, a certain type of substrate may not require all of the process steps as set up in a specific process plant, such as the plant 110 , whereby a different material flow or material flow is realized. It is assumed, for example, that the process steps 143 and 144 are not required for a certain type of product, so that a corresponding material flow has five steps, possibly in connection with corresponding sub-steps, as explained above. Similarly, if the chambers malfunction 113 , 114 or when these chambers are switched off due to maintenance and the like, a different process sequence can be set up, since the parallelism of the step 145 is no longer given. As a result, a current status of the process system can be determined by monitoring and evaluating the respective system messages and can be used to define a process flow as it is currently implemented in the system, while in other illustrative embodiments several material flows are recognized by, for example, historical data Processes are preserved that have been set up in the system under consideration over a certain period of time as described in more detail below. In addition, system events can be determined from the system messages in order to determine the respective time intervals for the various steps 141 , ..., 147 to determine, which includes corresponding sub-steps, such as the pre-process time interval, which represents a time in which the substrate is in the process chamber but is processed, a post-process time, ie a time interval after the actual processing of the substrate and before it is picked up by the transport system 155 as previously explained. It should be noted that corresponding process time intervals are determined on the basis of actually detected system events, with a large number of corresponding events being used for a special process time interval, whereby a high statistical relevance is achieved, which offers improved robustness with regard to random fluctuations.

1d shows schematically the manufacturing environment 150 according to an illustrative embodiment in which multiple process plants 110a , ..., 110c are provided, each of which has a structure similar to that previously with respect to the process plant 110 is explained. It should be noted, however, that in a typical manufacturing environment for producing semiconductor components, corresponding cluster systems have a higher degree of complexity than was previously the case with regard to the process system 110 is described. That is to say, a larger number of functional units is typically provided, some of which are operated on the basis of corresponding process recipes in order to set up a complex process sequence. In the in 1d The embodiment shown includes the manufacturing environment 150 furthermore corresponding system interfaces 116a , ..., 116c associated with the respective process equipment 110a , ..., 110c are connected, and are designed to provide system-specific information on a desired level of abstraction with a higher-level control system 151 exchange what an MES can represent, as is typically provided in a semiconductor manufacturing facility. It also includes the manufacturing environment 150 an automatic throughput model generator 100 , which has a data input section 101 has that functionally with the one or more system interfaces 116a , ..., 116c is connected to receive plant messages therefrom at a level of abstraction that reflects the plant events at the unit level, as previously discussed. In other words, in contrast to the data communication of the system interfaces 116a , ..., 116c with the higher-level control system 151 , which operates on the basis of state models, receives the data input section 101 the system messages, so that system events can be recognized for each individual functional unit in the corresponding process system 110a , ..., 110c to enable. Thus, the presence of corresponding functional units and their functional behavior, for example, with regard to the beginning of an activity and the ending of an activity by the data input section 101 be monitored. In some illustrative embodiments, is the data input section 101 further designed to store corresponding system messages, at least over a period of time or operating period, in order to enable statistical processing of the system messages with a desired degree of statistical significance.

Also includes the automatic throughput model generator 100 a model calculator 102 that works with the data input section 101 is connected and designed, one or more material flows or material flows in the systems 110 , ..., 110c in conjunction with associated process time intervals, as previously described. That is, during the operation of the throughput model generator 100 a corresponding system-related information is collected over a specified operating time, which is then stored and suitably preprocessed in order to thereby define the functional units of the process systems 110a , ..., 110c to determine. For example, during the specified period of operation, the respective system message activities can indicate a specific number of functional units in connection with corresponding time intervals which are required for the respective unit activities. The generator 100 can also recognize the parallelism of certain functional units by referring to the process recipes of the process chambers and / or by appropriately analyzing the system messages if, in a given cascade, only part of the substrates in a process chamber is processed while another part is processed in another process chamber for otherwise identical process steps. Furthermore, the assigned process time intervals can be based on an event recording within the module 101 can be determined, whereby the identification of one or more "limiting" resources is possible, ie functional units that actually cover the interval PI (see 1c ), since this represents the units that represent the lowest capacity for the current operating status of the process plant under consideration. Furthermore, recipe-dependent process time intervals and also minimal static process time intervals can be set on the basis of corresponding event recording properties of the data input section 101 to be determined.

1e shows schematically the manufacturing environment 150 with the throughput model generator 100 according to even more vividly Embodiments. As shown, includes the environment 150 the multiple process plants 110a , ..., 110c that are functional with the generator 100 are connected in order to provide the system messages. In the embodiment shown, the process plants have 110a , ..., 110c built-in suitable interfaces for data transfer to the generator 100 without the need for suitably designed system interfaces that can also ensure data exchange with a higher-level control system. The moreover one is the generator 100 designed to receive additional system-specific or process-specific information from an external source, such as the MES 151 or another suitable source of information. This means that relevant information, such as recipes, can be saved in the process systems 110a , ..., 110c and the like from the external source 151 can be obtained if the relevant information is not received directly from the process equipment 110a , ..., 110c is available. In the illustrated embodiment, the throughput model generator comprises 100 the data input section 101 for receiving the system messages in the form of unprocessed data which, in some illustrative embodiments, can be stored at least for a certain period of time if the real-time processing of the data is not available. The generator 100 comprises a data preprocessing module 103 , which has set up therein suitable algorithms in order to extract relevant information from the system messages, for example for the recognition of functional units and also the recognition of corresponding start times and end times of activities. Furthermore, the data preprocessing module 103 have statistical analysis algorithms set up therein in order to carry out processing on the basis of the process time intervals which were extracted from the system messages in order to determine representative process time intervals for the corresponding steps of a material flow which is characterized by the system messages.

For example, recipe-dependent process time intervals, as shown in the steps 143 and 145 in 1b are involved, by an average of a plurality of corresponding process time intervals is determined for each of these steps and the corresponding average is used as a representative process time interval for the respective process step. Similarly, suitable process time intervals for other unit activities that do not depend on process recipes can be determined, as described in more detail below. The throughput module generator 100 further includes, in one illustrative embodiment, a database to maintain a data model associated with the one or more process tools 110a , ..., 110c is equivalent to. For example, the database is 104 functional with the data preprocessing module 103 connected to store corresponding database entries showing the material flow and corresponding representative process time intervals used by the data preprocessing module 103 be determined, concern. Furthermore, a modeling module 105 provided so that this works with the database 104 is connected in order to retrieve a suitable set of data therefrom, as is required for a special process plant and a specified process scenario. That is, the modeling module 105 can for example on request by means of a user interface 106 retrieve an appropriate set of data representing material flow associated with associated representative process time intervals. Thus, by generating the corresponding throughput model, the module 105 the suitable platform for a corresponding process scenario of the one or more systems 110a , ..., 110c , for example at the request of a user via the user interface 106 , offer on the basis of which the model calculator 102 actually determines a throughput value, for example on the basis of user-specified input parameters, such as the length of the cascade and the like. Corresponding throughput results can thus be obtained via the user interface 106 may be provided, wherein in some illustrative embodiments, in addition to the corresponding throughput-related parameters for the module 105 and the model calculator 102 the user interface also with the database 104 and / or the data preprocessing module 103 and / or the data input section 101 communicates. For example, raw data can be sent via the user interface 106 can be observed, with a suitable preselection of raw data by interaction with the data input section 101 via the user interface 106 can be effected. Intermediate results can be generated in a similar way, for example after the raw data has been preprocessed via the user interface 106 when acting on the data preprocessing module 103 to be viewed as. In a similar way one can access the database 104 via the user interface 106 take place, with corresponding rights to access one or more of the modules of the generator 100 with a hierarchy level for users who use the user interface 106 operate, can be linked. For example, access to the database 104 , for example for modifying data and / or for interacting with the data preprocessing module, for example with regard to modifying statistical analysis techniques and the like, should only be restricted to privileged users.

In one illustrative embodiment, there is a test unit 107 provided to the validity verify the data generated by the preprocessing module based on attachment messages received from the data input section 101 can be obtained as described in more detail below.

During the operation of the throughput model generator 100 like him in 1e shown, raw data is added to the data input section 101 which contain, for example, basic setting events for machine interfaces or directly from an in-plant control system, and / or process-specific data and other plant data, such as messages about units that are configured in the MES, and the like can be used. In some illustrative embodiments, the data in section receives 101 continuous corresponding data and saves the raw data for further processing by the data preprocessing module 103 in an appropriate manner. In other cases, corresponding raw data are only available for specified operating times depending on the resources of the data input section 101 fed. The data preprocessing module 103 determines the data records for the data model in the database 104 for example, on a regular basis or at any suitable time, or in some illustrative embodiments essentially on a continuous basis, thereby creating corresponding entries for the database 104 be generated. The entries in the database 104 can also be configured to monitor the history of the throughput-related data, which can be advantageous in terms of identifying material flows, as described below, and also in terms of estimating changes in throughput over time.

1f shows schematically a structure of the data model in the database 104 according to illustrative embodiments. The data model represents one or more process systems that are distributed in corresponding “sub-items” or functional units, at least some of which are linked to corresponding process recipes. In other words, as explained above, some of the functional units, such as process chambers, are operated on the basis of an assigned process recipe, which, however, can change depending on the product type, the results of research and development activities and the like. The data model thus represents the “hardware” of the system under consideration in the form of logical units “system”, “sub-item” and “recipes”. Furthermore, the corresponding functional behavior, ie the “software” of the systems can be represented logically by several material flows that are created on the basis of the functional units in accordance with the respective production processes for the products under consideration. The respective flows can be represented by the number and a corresponding sequence of steps, ie by activities that are carried out by the respective functional units. Thus, some of the steps are also linked to process recipes, such as this through 1f is specified. Furthermore, corresponding steps of each of the flows are linked to corresponding process time intervals “step times”, which can depend on recipes, depending on whether the respective steps linked to a functional unit are operated on the basis of an assigned process recipe. Thus, the structure of the data model is the "hardware" and the "software" of each process plant in the database 104 appropriately designed to build a throughput model based on a desired operating scenario.

In other words, if necessary, the configuration of a process plant can be "modified" by selecting a suitable process sequence with the corresponding step times in connection with the assigned recipes. For example, as previously with reference to the 1a until 1c is explained, a process chamber can be “switched off” in the throughput model by selecting a corresponding material flow in which, for example, one of the two parallel process chambers is no longer present. Thus, in connection with the process time intervals associated with this particular process, the throughput can then be calculated on the basis of a model which is built up from the particular parameters which reflect the operating status under consideration. If, in a similar way, a change to a recipe is to be evaluated with regard to changes in the throughput of a process plant under consideration for otherwise identical conditions, ie the same material flow is considered, suitable process time intervals can be selected depending on the recipe from the database 104 can be accessed. It should be noted that a corresponding logical structure can be preserved for different time periods, which enables efficient monitoring of throughput over time for a large number of different process scenarios. Furthermore, corresponding material flows that were previously through the data preprocessing module 103 recognized in the database 104 and can be used for flow detection in a later phase, for example by comparing a currently identified material flow with several previously determined material flows in order to identify the same material flows or to identify a similarity between an earlier material flow and a currently identified material flow.

1g shows schematically a decision tree in the data preprocessing module 103 may be arranged to efficiently flow material based on the from the data input section 101 to recognize received system messages. That is, to define the basic process steps in a material flow, a large number of decisions are made automatically in order to determine the process steps in order to determine the process steps for the statistical samples of the assigned time intervals from the data input section 101 to obtain. The in 1g The decision tree shown is an abstract example for setting up the detection of a material flow on the basis of an automated procedure. As shown, in stage 1 of the decision tree the number of steps is determined on the basis of the system message, ie the various activities of the respective functional units for the complete processing of a substrate in the process system are determined. For example, the data preprocessing module takes effect 103 on the database 104 to determine whether a material flow with the same number of steps as determined in a previous period is stored. If a material flow with the same number of steps is not in the database 104 is stored, then the database decides 104 that a new material flow has been recognized and the corresponding sequence of process steps, which is recognized from the system messages, is added to the database 104 as a new material flow. If a material flow with the same number of steps is recognized in stage 1, the decision recognition algorithm goes on to stage 2, in which it is decided whether an identical material flow is in the database 104 was saved. For example, it can be determined whether the same functional units are involved in the previous material flow and the currently recognized material flow. Furthermore, the sequence of the process steps can be compared with the sequence of process steps of the material flow contained in the data model. If, in stage 2, the currently recognized material flow is assessed as being identical to a previously stored material flow, corresponding process time intervals can, however, be added to the previously stored material flow with updated time stamps if necessary. If it is decided in stage 2 that a currently recognized material flow is different, then the decision-making algorithm proceeds to stage 3, in which it is decided whether the current material flow is "very similar" to a previously stored material flow. If it is decided in stage 3 that the material flows are not very similar, it is decided that a new material flow has been recognized. If, in stage 3, the currently recognized material flow is classified as very similar to a previous material flow, a modified flow of the previously stored material flow can be set up and can be identified as a modified version. For example, a very similar flow can be represented as a material flow in which the parallelism of one or more steps is reduced or increased due to the separation of a functional unit or due to the addition of a further functional unit. It should be noted that additional or different stages are in the decision-making space of the 1g can be set up, whereby this depends on the desired “resolution” when differentiating between material flows in a special process plant.

After the material flow has been identified, the context of this flow is used to determine statistical samples of various required time intervals that are related to the functional units linked to the identified material flow. This means that the system messages are examined with regard to system events that relate to the same material flow steps in order to determine relevant time interval values.

According to illustrative embodiments, a distinction is made between corresponding process time intervals in order to obtain two types of time intervals, for example static time intervals and recipe-dependent time intervals. Static time intervals, such as window replacement times, post-process time intervals, d. H. the actual process is finished but the substrate is still in the process chamber. H. the process has not yet started but the substrate is in the process chamber and linked to a single functional unit of the process system. Thus, the time interval required to complete these steps does not depend on the recipe selected. In contrast to this, recipe-dependent process time intervals also change according to the selected process recipe. The length of the static time intervals and the recipe-dependent time intervals is defined by the length of time between two predefined recording events that are transmitted via the system messages and are extracted from them by analyzing the data packets or another information structure or format as it is used to transmit the system messages will. For example, the messages can be transmitted based on standard protocols as previously described.

1h shows schematically a time diagram for the processing of substrates in a functional unit, for example a process chamber on the basis of a specified process recipe, a corresponding recipe-dependent process time interval being referred to as ETp (Entity Time process). Accordingly, there are corresponding pre-process and post-process time intervals as ETpre and ETpost denotes and represent time intervals in which the substrate is present in the respective functional unit or process chamber without actual processing taking place. Furthermore, a time interval for exchanging the substrate in the process chamber under consideration is specified as ETs (Entity Time swap). Thus, as shown, a sequence of activities 160 begin with the preprocessing time ETpre, which is followed by the actual processing of a substrate ETp. After the end of the actual process, the substrate remains in the chamber, as indicated by the interval ETpost, after which the substrate is removed and another substrate is placed in the chamber, the corresponding substrate handling activities being indicated by the time interval ETs. A further sequence of pre-process time ETpre, actual process time ETp and post-process time ETpost can then follow in the chamber under consideration.

The portion of the diagram indicated as 161 represents a sequence of recording events associated with the sequence 160 is correlated. Thus, at the beginning of the preprocessing time, a "target at" event is generated and is transmitted via a system message. When the substrate is actually processed, a “process started” event is generated and a “process ended” event is generated at the end and ETp. When the substrate is removed from the process chamber under consideration, a “slice of” event is generated and after another substrate arrives in the process chamber under consideration, a further “slice of” event is generated. Consequently, based on the appropriate recording events 161 the ETp time intervals for the material flow under consideration and the corresponding process recipe can be determined by extracting the events from the system messages. It should be noted that, depending on the process and the recipe, the times ETp can vary, for example due to different end points that are recognized by a corresponding mechanism and the like. In other cases, typical random fluctuations can occur for otherwise identical process conditions, which also contributes to a different recipe-dependent process time ETp. Thus, on the basis of a large number of these recipe-dependent time intervals, a corresponding representative, recipe-dependent time interval can be determined by, for example, calculating a suitable average value. The corresponding representative process time interval can be determined for each of the recipe-dependent process steps of the material flow under consideration, whereby a representative total process time of a substrate is determined in relation to the material flow under consideration. It should be noted that a representative recipe-dependent process time interval can be determined on the basis of any statistical technique, for example by eliminating outliers, by fitting a suitable curve and the like, this being done by the computing resources of the data preprocessing module 103 depends.

With reference to the 1i until 1k the determination of process time intervals is described in greater detail in accordance with illustrative embodiments that are not dependent on a recipe. For the calculation of static time intervals, the functional status of a unit is taken into account by “considering” events for substrates that are processed in the respective unit. However, there can be a complex mutual dependency between the various functional units, which can influence a corresponding process time interval.

1i shows schematically a timing diagram 163 , in which the activities of a first functional unit, such as a robot transport system, and a second functional unit, such as a process chamber are shown. In the example shown, the robot transport system has an “actual” process time ETp and a post-process time ETpost, which represent the real minimum operating time of the robot transport system. As shown, this combined time interval can be longer than an actual process time of the process chamber, whereby an additional waiting time of the process chamber is required after the end of the actual process in the process chamber, whereby a "wrong" post-process time "ETpost" is generated for the process chamber.

1y shows schematically a situation in which the overall behavior of the robot transport system and the chamber loaded by the robot transport system is limited by the process that is carried out in the process chamber. Thus, the actual process time of the chamber ETp and the required post-process time ETpost can be significantly longer compared to the actual process time of the robot transport system ETp, whereby an incorrect process time “ETpost” is generated for the robot transport system. Consequently, the extraction of the respective post-process times according to the in the 1i and 1y situations described based on facility messages conveying an event recording sequence, as in the sequence 161 shown, lead to an overestimate of the required post-process time, which can lead to significantly different throughput values if the throughput is calculated on the basis of, for example, a modified process recipe while an essentially unmodified material flow is used.

1k shows schematically the determination of suitable static process time intervals, with essentially a negative influence of mutual dependency, as described with reference to the 1i and 1y is avoided. 1k shows schematically the distribution of process time intervals for a special functional unit for carrying out its assigned activity that is not dependent on a process recipe, for example on a robot transport system and the like. In order to achieve a high degree of robustness in the throughput model to be created, a reasonable minimum process interval, ie a time interval that actually describes the "real" mechanical capacity of the unit under consideration, is determined on the basis of a moderately large number of time intervals that are determined by analysis many recording events are actually determined, for example by events as described in 1h are shown. In an illustrative embodiment, this is achieved on the basis of percentiles or exceedance probabilities, which is understood as a mathematical algorithm in order to determine a threshold value above or below which a specified percentage of events of a distribution is arranged. For example, in the illustrative example shown, a five percentile is used that represents the time interval T under which 5% of the distribution are arranged, while the remaining 95% of the recognized process time intervals have a value greater than the threshold value T. It should be noted, however, that other percentile values can be used if desired. Furthermore, other statistical rules can be used to determine a suitable “minimum” process time, which can be used as a representative process time interval for the recipe-independent functional unit under consideration.

Thus, the data preprocessing module 103 automatically determine corresponding process time intervals for recipe-dependent units and / or recipe-independent units and can add the corresponding representative time intervals in connection with a corresponding material flow to the database 104 in order to determine the current "hardware" and "software" status of the considered situation on the basis of solid experimental data.

It should be noted that system events extracted from the system messages, for example, in the form of recording events, as shown in 1h represent an efficient means of determining material flows and the time intervals. In some illustrative embodiments, a higher degree of robustness and accuracy is achieved by verifying the admissibility of the time sequence of corresponding system events in order to discard system events that have not been verified. That is, the sequence of recording events 161 (please refer 1h ) can be verified with regard to their authenticity in order to essentially exclude erroneously recognized sequences. Thus, according to illustrative embodiments, a test unit can be provided, for example in the form of the test unit 107 as in 1e is shown. In an illustrative embodiment, the test unit 107 set up to verify the validity of the temporal sequence of recording events in the form of a state machine, which can be a freely configurable state machine for defining permissible event transitions. Thus, the time sequence of recording events is verified by applying the sequence of the state machine to verify whether a valid sequence of recording events has been mediated, also indicating whether or not the respective time intervals are valid candidates for determining corresponding process time intervals.

1l Figure 3 shows schematically an example of the state machine representing a temporal sequence of recording events. For example, represents a state S1 a "disk has started" event, which is a condition S2 "Disk at" connects. According to the state machine shown, the state follows S2 a sequence of states S3 and S4 "Transport started" and "Transport finished" accordingly, ie the arrival of a substrate is linked to the beginning and the end of a transport activity. Similarly, a sequence of states follows S5 and S6 on the state S2 , which represent "process started" and "process finished", ie according to the state S2 , which corresponds to the arrival of the substrate at a certain functional unit, a “process started” must be followed by a “process completed”. Similarly, according to the state must S2 and the corresponding possible branching into the sequences S3 , S4 or S5 , S6 a condition S7 "Slice of" from the state S8 "Disc finished" be accompanied. When verifying the sequences based on the recording events 161 be detached (see 1h ) a verification of the time sequence and thus the validity of the corresponding process time intervals can be achieved. Thus, the overall accuracy of the corresponding throughput model can be improved.

As previously explained, the data model in the database 104 Contain entries according to a logical structure as previously referring to 1f is explained. Furthermore, in some illustrative embodiments, a history of throughput-related data is established on the basis of corresponding database entries, for example in order to recognize long-term changes in throughput or to enable throughput results to be determined on the basis of different operating phases for any desired process scenario.

1m shows schematically corresponding database entries or points which have suitable time stamps in order to enable the time context of the corresponding entries to be recognized. For example, contains a "static" database element 104a a time stamp 104i indicating the time of the first appearance of the item 104a indicates and contains a second timestamp 104f showing the time of the last appearance of the item 104a indicates. For example, the element 104a represent a "material flow" that is only once in the database 104 shows up. Thus, by adding the brands 104i , 104f the history of corresponding elements 104a be monitored. The database also includes 104 elements 104b , which can contain stored time intervals, and which for each new process flow of the throughput model generator 100 be generated. Thus, by adding a "runtime identifier" 104g , which also indicates a time context of the run under consideration, a history of corresponding calculated process time intervals can be obtained, which can be advantageous in order to carry out successive throughput calculations, which are therefore carried out on the basis of said times that are used for each run of the throughput model generator 100 were determined.

The following therefore applies: The present disclosure provides systems and methods for estimating the throughput of cluster systems in a manufacturing environment on the basis of automatically generated throughput models. For this purpose, system-specific messages are collected and analyzed in order to identify material flows and corresponding process time intervals with high statistical relevance. As a result, throughput-related information can be generated for a large number of process scenarios with a high degree of robustness against random fluctuations and with a high degree of consistency.

Further modifications and variations of the present disclosure will become apparent to those skilled in the art in light of this description. Therefore, this description is intended to be illustrative only and for the purpose of providing one skilled in the art with the general manner of practicing the principles disclosed herein. It should be understood that the forms shown and described herein are to be regarded as the presently preferred embodiments.

Claims (15)

Throughput model generator (100) having: a data input section (101) which is designed to receive system-specific information about the configuration and operating status of a process system (110, 110A, 110B, 110C) that is used in a production environment (150), the process system ( 110, 110A, 110B, 110C) comprises a plurality of functional units (111, .., 115) and wherein the plant-specific information comprises plant events that are generated by the plurality of functional units (111, .., 115) of the process plant (110, 110A, 110B, 110C ) be generated; a database (104) which is functionally connected to the data input section (101) and is designed to store a data model of the process plant (110, 110A, 110B, 110C), the data model showing at least one material flow in the process plant (110, 110A, 11 OB, 110C); a model calculator (102) which is functionally connected to the database (104) and is designed to determine a throughput of the process plant (110, 110A, 110B, 110C) on the basis of the data model and one or more throughput-related parameters; a user interface (106) operatively connected to the model calculator (102) and configured to select a value of the one or more throughput-related parameters; a data preprocessing module (103) which is functionally connected to the data input section (101) and the database (104) and is designed to determine process steps and associated process times for the material flow in the process plant (110, 110A, 11OB, 110C), the The data preprocessing module (103) is further designed to determine the process time of each process step by recognizing a plurality of system events for each process step, which is contained in the system-specific information, and by determining a representative time interval for each process step, and wherein the data preprocessing module (103) is further designed , a first representative time interval for a first process step that is not is related to a process recipe, and to determine a second representative time interval for a second process step related to a process recipe; and wherein the model calculator (102) is further configured to determine the throughput of the process plant (110, 110A, 110B, 110C) on the basis of the data model in relation to the first and the second representative time interval. Throughput model generator according to Claim 1 , in which the data preprocessing module (103) is further designed to display the first representative time interval for the first process step, which is not related to a process recipe, and the second representative time interval for the second process step, which is related to a process recipe based on a percentile of time intervals related to the first and second process steps. Throughput model generator according to Claim 1 which also has a test unit (107) which is connected to the data preprocessing module (103) and is designed to validate a time sequence of the multiple system events. Throughput model generator (100) according to Claim 3 wherein the test unit (107) has a state machine which represents all system events in the process system (110, 110A, 110B, 110C) as state transitions. Throughput model generator (100) according to Claim 4 which further comprises a plant event database which is connected to the process plant (110, 110A, 110B, 110C) and the input section and is configured to receive and store plant messages of a process plant (110, 110A, 110B, 110C). Throughput model generator (100) according to Claim 1 , wherein the process plant (110, 110A, 110B, 110C) is a process plant (110, 110A, 110B, 110C) which has a plurality of functional units (111, .., 115), each of which is a subset of the plurality of functional units (111 ,. ., 115) is designed to process a substrate (130) according to a specified process recipe and wherein each of a second subset of the plurality of functional units (111, .., 115) is designed to carry out substrate transport activities. A method for automatically determining a throughput of a process plant (110, 110A, 110B, 110C), the method comprising: Recognizing all functional units (111, .., 115) of the process plant (110, 110A, 110B, 110C) by monitoring plant messages that are generated by the process plant (110, 110A, 110B, 110C) during operation; Recognizing one or more material flows as a sequence of process steps which are carried out by at least some of the functional units (111, .., 115); Determining a process time interval for each of the process steps of the one of the plurality of material flows; and Determining a throughput of the process plant (110, 110A, 110B, 110C) for a selected process scenario for the one or more material flows using the determined process time intervals, wherein determining a process time interval for each of the process steps comprises: Detection of first functional units (111, .., 115) that are not dependent on an assigned process recipe, and detection of second functional units (111, .., 115) that are dependent on an assigned process recipe, the first and second Functional units (111, .., 115) are recognized using at least the system messages; and wherein the process interval is determined on the basis of the recognized first functional units (111, .., 115) and second functional units (111, .., 115). Procedure according to Claim 7 wherein determining a process time interval for each of the process steps comprises: statistically evaluating a plurality of process time intervals which correspond to a specific one of the process steps in order to obtain a representative process time interval for the specific process step. Procedure according to Claim 8 wherein obtaining the representative process time interval of one of the first functional units (111, .., 115) comprises: determining a distribution of the plurality of process times and selecting a threshold value in the distribution as the representative process time interval. Procedure according to Claim 9 , wherein the threshold is selected based on a specified percentile of the distribution. Procedure according to Claim 7 further comprising: storing data representing the one or more material flows and process time intervals in a database (104). Procedure according to Claim 11 further comprising: updating the database (104) on a regular basis while preserving at least some of the data previously fed to the database (104) to preserve a history of the one or more material flows and process time intervals. Procedure according to Claim 12 wherein the one or more material flows are determined by comparing a material flow determined using the facility messages with a history of the one or more material flows represented in the database (104). Procedure according to Claim 7 which further comprises: verifying the validity of a time sequence of the attachment messages and using the attachment messages with a successfully verified validity to determine the process time intervals. Procedure according to Claim 14 , whereby the validity of the time sequence is verified by a state machine that handles all plant events of the process plant (110, 110A, 110B, 110C) is represented.

Images