... Autores: José Belarmino Pulido Junquera; Directores de la Tesis: Carlos Javier Alonso Gonzále... more ... Autores: José Belarmino Pulido Junquera; Directores de la Tesis: Carlos Javier Alonso González; Lectura: En la Universidad de Valladolid ( España ) en 2001; ... José Manuel Marqués Corral ( voc. ); UNESCO : 12 Matemáticas: 1203 Ciencia de los ordenadores: 120304 ...
In this chapter, several application examples are introduced, which will be used to illustrate th... more In this chapter, several application examples are introduced, which will be used to illustrate the diagnosis schemes presented and studied in this book. The objective of introducing these application examples in the first part of this book, immediately after the introduction, is to provide the reader with the useful application background and understandings of some basic technical concepts in the process monitoring and fault diagnosis field.
In this chapter, the artificial intelligence approach to model-based diagnosis is introduced. Fir... more In this chapter, the artificial intelligence approach to model-based diagnosis is introduced. First, we present the main ideas of the Consistency-Based Diagnosis (CBD) methodology (the no-function-in-structure principle, the use of models of correct behavior, and the requirement of local propagation in the models), together with its logical formalization provided by Reiter’s work. In CBD, concepts such as (minimal) conflicts and (minimal) diagnoses play a major role because they allow to characterize and to compute the whole set of diagnosis in an automated way. Second, we introduce the General Diagnosis Engine (GDE) which is the de facto computational paradigm for CBD, and we explain how it works. Finally, to increase the discriminative power in CBD results due to using only correct behavior models, we introduce the concept of fault models and explain how CBD can be extended to with predictive fault models, while retaining the essential no exoneration assumption.
This paper proposes a diagnosis architecture that integrates consistency based diagnosis with ind... more This paper proposes a diagnosis architecture that integrates consistency based diagnosis with induced time series classifiers, trying to combine the advantages of both methods. Consistency based diagnosis allows fault detection and localization without prior knowledge of the device fault modes. Machine learning techniques are able to induce time series classifiers that may be used to identify fault modes of a dynamic systems. The diagnostician performs fault detection and localization resorting to consistency based diagnosis through possible conflicts. Then, a time series classifier, induced from simulated examples, generates a sequence of faults modes, coherent with the result of the fault localization stage, and ordered by fault modes confidence. Finally, to simplify the diagnosis task, it is considered as a subtask of a supervisory system, who is in charge of identifying the working conditions for the physical system.
This work is focused on structural approaches to studying diagnosability properties given a syste... more This work is focused on structural approaches to studying diagnosability properties given a system model taking into account, both simultaneously or separately, integral and differential causal interpretations for differential constraints. We develop a model characterization and corresponding algorithms, for studying system diagnosability using a structural decomposition that avoids generating the full set of system ARRs. Simultaneous application of integral and differential causal interpretations for differential constraints results in a mixed causality interpretation for the system. The added power of mixed causality is demonstrated using a case study. Finally, we summarize our work and provide a discussion of the advantages of mixed causality over just derivative or just integral causality.
The FDI and DX communities have developed complementary approaches that exploit structural relati... more The FDI and DX communities have developed complementary approaches that exploit structural relations in the system model to find efficient solutions for the residual generation and residual evaluation steps in fault detection and isolation in dynamic systems. This paper compares three different structural techniques, two from the DX community and one from the FDI community. To simplify our comparison, we start with a common modeling approach that employs bond graphs. We describe the residual generation methods used by the three approaches, and apply them to a standard three tank configuration to demonstrate their diagnostic ability for continuous, nonlinear systems. 1.
This work introduces DXPCS, a software tool performing modelbased diagnosis of continuous dynamic... more This work introduces DXPCS, a software tool performing modelbased diagnosis of continuous dynamic systems whose models can be represented as a set of Algebraric/Ordinary Differential Equations. The diagnosis approach implemented is based upon the Possible Conflict (PC) concept. DXPCS is mainly intended for educational purposes, providing a complete package to show the Artificial Intelligence approach to model-based diagnosis for postgraduate students. Given a set of equations, together with structural information about the model, DXPCS is able to automatically build simulation models for each PC, it can handle both single-fault and multiple-fault scenarios, for both parametric and additive faults. Different options for fault detection, residual generation and evaluation can be chosen. The software architecture and the software performance for one simple case study, are provided in this paper.
Consistency-based diagnosis of dynamic systems using possible conflicts rely upon a semi-closed l... more Consistency-based diagnosis of dynamic systems using possible conflicts rely upon a semi-closed loop simulation of numerical models. Simulation approaches need to know the initial state, which is a nontrivial requirement in real-world systems. Prognosis approaches also require techniques for predicting the future system states under nominal and faulty conditions. This work proposes to integrate state observers to estimate initial states for simulation within the consistency-based diagnosis framework using possible conflicts. This work extends the BRIDGE framework for one class of dynamic systems, using the possible conflict concept to find every subsystem with necessary structural redundancy to lead to a minimal conflict activation. These algorithms can analyze those structures, without additional information, and point out possible implementations as observers or sim-
This paper propose a diagnosis architecture that integrates consistency based diagnosis with indu... more This paper propose a diagnosis architecture that integrates consistency based diagnosis with induced time series classi ers, trying to combine the advantages of both methods. Consistency based diagnosis allows fault detection and localization without prior knowledge of the device fault modes. Machine learning techniques are able to induce time series classi ers that may be used to identify fault modes of a dynamic systems. The diagnostician performs fault detection and localization resorting to consistency based diagnosis trough possible conicts. Then, a time series classi er, induced from simulated examples, generates a sequence of faults modes, coherent with the result of the fault localization stage, and ordered by fault modes con dence.
Accurate modeling mechanisms play an important role in model-based diagnosis, and the bond graph ... more Accurate modeling mechanisms play an important role in model-based diagnosis, and the bond graph modeling language has proved to be helpful for this task. In this paper we present an algorithm for automatically derive ARR-like structures, possible conflicts, from the bond graph model of a system. The algorithm uses temporal causal graphs as an intermediate structure to generate the set of possible conflicts. Performance of the algorithm for structural and sensor faults is then studied. Finally, we present another algorithm to automatically derive temporal information in the fault signature matrix for the set of possible conflicts, thus improving the isolation capabilities of the approach.
Complex hybrid systems are present in a large range of engineering applications, like mechanical ... more Complex hybrid systems are present in a large range of engineering applications, like mechanical systems, electrical circuits, or embedded computation systems. The behavior of these systems is made up of continuous and discrete event dynamics that increase the difficulties for accurate and timely online fault diagnosis. The Hybrid Diagnosis Engine (HyDE) offers flexibility to the diagnosis application designer to choose the modeling paradigm and the reasoning algorithms. The HyDE architecture supports the use of multiple modeling paradigms at the component and system level. However, HyDE faces some problems regarding performance in terms of complexity and time. Our focus in this paper is on developing efficient model-based methodologies for online fault diagnosis in complex hybrid systems. To do this, we propose a diagnosis framework where structural model decomposition is integrated within the HyDE diagnosis framework to reduce the computational complexity associated with the fault...
In this chapter, we analyze main problems found by the Artificial Intelligence approach to Model-... more In this chapter, we analyze main problems found by the Artificial Intelligence approach to Model-based diagnosis (DX): the online computation of minimal conflicts by means of an ATMS-like dependency-recording engine, and the need for an extension to deal with dynamic systems diagnosis. To cope with the first problem we will see different options: from extensions to the original GDE to the description of several topological methods, explaining deeply one of them: the Possible Conflict (PC) approach, and its relation with minimal conflicts and ARRs. To cope with the second problem, dynamics, we review the whole set of proposals made to extend Reiter’s formalization and the GDE to dynamic systems: from GDE extensions to the natural extension of topological methods to include temporal information. In this chapter we provide the complete extension of the PCs approach to diagnose dynamic systems, and their relation not only with ARRs, but with another FDI proposals for systems tracking: s...
In this work we introduce DXPCS, a software tool capable of performing consistency-based diagnosi... more In this work we introduce DXPCS, a software tool capable of performing consistency-based diagnosis of continuous dynamic systems whose models can be represented as a set of Ordinary Differential Equations. The diagnosis approach relies upon the Possible Conflict, PC for short, concept. DXPCS is able to automatically build the simulation models for each PC. Single-fault and multiple-fault scenarios, for both parametric and additive faults, can be injected, and studied. DXPCS allows the integration of different algorithms for fault detection, residual generation and evaluation, together with an incremental version of the minimal-hitting set algorithm for fault localization. The software architecture, together with performance results for one simple case study, are provided in this paper.
Reliable and timely fault detection and isolation are necessary tasks to guarantee continuous per... more Reliable and timely fault detection and isolation are necessary tasks to guarantee continuous performance in complex industrial systems, avoiding failure propagation in the system and helping to minimize downtime. Model-based diagnosis fulfils those requirements, and has the additional advantage of using reusable models. However, reusing existing complex non-linear models for diagnosis in large industrial systems is not straightforward. Most of the times the models have been created for other purposes different from diagnosis, and many times the required analytical redundancy is small. In this work we propose to use Possible Conflicts, which is a model decomposition technique, to provide the structure (equations, inputs, outputs, and state variables) of minimal models able to perform fault detection and isolation. Such structural information can be used to design a gray box model by means of state space neural networks. We demonstrate the feasibility of the approach in an evaporator...
Fault Diagnosis of Hybrid Dynamic and Complex Systems
Nowadays hybrid systems are everywhere: vehicles, planes, electronic devices, industrial factorie... more Nowadays hybrid systems are everywhere: vehicles, planes, electronic devices, industrial factories, and so on. All these systems exhibit different behavior patterns depending on the actual operation mode. In this work we propose a framework for fault diagnosis of those dynamic systems characterized by continuous behavior commanded by discrete actuators such as valves, bypasses, relays, etc.
... Autores: José Belarmino Pulido Junquera; Directores de la Tesis: Carlos Javier Alonso Gonzále... more ... Autores: José Belarmino Pulido Junquera; Directores de la Tesis: Carlos Javier Alonso González; Lectura: En la Universidad de Valladolid ( España ) en 2001; ... José Manuel Marqués Corral ( voc. ); UNESCO : 12 Matemáticas: 1203 Ciencia de los ordenadores: 120304 ...
In this chapter, several application examples are introduced, which will be used to illustrate th... more In this chapter, several application examples are introduced, which will be used to illustrate the diagnosis schemes presented and studied in this book. The objective of introducing these application examples in the first part of this book, immediately after the introduction, is to provide the reader with the useful application background and understandings of some basic technical concepts in the process monitoring and fault diagnosis field.
In this chapter, the artificial intelligence approach to model-based diagnosis is introduced. Fir... more In this chapter, the artificial intelligence approach to model-based diagnosis is introduced. First, we present the main ideas of the Consistency-Based Diagnosis (CBD) methodology (the no-function-in-structure principle, the use of models of correct behavior, and the requirement of local propagation in the models), together with its logical formalization provided by Reiter’s work. In CBD, concepts such as (minimal) conflicts and (minimal) diagnoses play a major role because they allow to characterize and to compute the whole set of diagnosis in an automated way. Second, we introduce the General Diagnosis Engine (GDE) which is the de facto computational paradigm for CBD, and we explain how it works. Finally, to increase the discriminative power in CBD results due to using only correct behavior models, we introduce the concept of fault models and explain how CBD can be extended to with predictive fault models, while retaining the essential no exoneration assumption.
This paper proposes a diagnosis architecture that integrates consistency based diagnosis with ind... more This paper proposes a diagnosis architecture that integrates consistency based diagnosis with induced time series classifiers, trying to combine the advantages of both methods. Consistency based diagnosis allows fault detection and localization without prior knowledge of the device fault modes. Machine learning techniques are able to induce time series classifiers that may be used to identify fault modes of a dynamic systems. The diagnostician performs fault detection and localization resorting to consistency based diagnosis through possible conflicts. Then, a time series classifier, induced from simulated examples, generates a sequence of faults modes, coherent with the result of the fault localization stage, and ordered by fault modes confidence. Finally, to simplify the diagnosis task, it is considered as a subtask of a supervisory system, who is in charge of identifying the working conditions for the physical system.
This work is focused on structural approaches to studying diagnosability properties given a syste... more This work is focused on structural approaches to studying diagnosability properties given a system model taking into account, both simultaneously or separately, integral and differential causal interpretations for differential constraints. We develop a model characterization and corresponding algorithms, for studying system diagnosability using a structural decomposition that avoids generating the full set of system ARRs. Simultaneous application of integral and differential causal interpretations for differential constraints results in a mixed causality interpretation for the system. The added power of mixed causality is demonstrated using a case study. Finally, we summarize our work and provide a discussion of the advantages of mixed causality over just derivative or just integral causality.
The FDI and DX communities have developed complementary approaches that exploit structural relati... more The FDI and DX communities have developed complementary approaches that exploit structural relations in the system model to find efficient solutions for the residual generation and residual evaluation steps in fault detection and isolation in dynamic systems. This paper compares three different structural techniques, two from the DX community and one from the FDI community. To simplify our comparison, we start with a common modeling approach that employs bond graphs. We describe the residual generation methods used by the three approaches, and apply them to a standard three tank configuration to demonstrate their diagnostic ability for continuous, nonlinear systems. 1.
This work introduces DXPCS, a software tool performing modelbased diagnosis of continuous dynamic... more This work introduces DXPCS, a software tool performing modelbased diagnosis of continuous dynamic systems whose models can be represented as a set of Algebraric/Ordinary Differential Equations. The diagnosis approach implemented is based upon the Possible Conflict (PC) concept. DXPCS is mainly intended for educational purposes, providing a complete package to show the Artificial Intelligence approach to model-based diagnosis for postgraduate students. Given a set of equations, together with structural information about the model, DXPCS is able to automatically build simulation models for each PC, it can handle both single-fault and multiple-fault scenarios, for both parametric and additive faults. Different options for fault detection, residual generation and evaluation can be chosen. The software architecture and the software performance for one simple case study, are provided in this paper.
Consistency-based diagnosis of dynamic systems using possible conflicts rely upon a semi-closed l... more Consistency-based diagnosis of dynamic systems using possible conflicts rely upon a semi-closed loop simulation of numerical models. Simulation approaches need to know the initial state, which is a nontrivial requirement in real-world systems. Prognosis approaches also require techniques for predicting the future system states under nominal and faulty conditions. This work proposes to integrate state observers to estimate initial states for simulation within the consistency-based diagnosis framework using possible conflicts. This work extends the BRIDGE framework for one class of dynamic systems, using the possible conflict concept to find every subsystem with necessary structural redundancy to lead to a minimal conflict activation. These algorithms can analyze those structures, without additional information, and point out possible implementations as observers or sim-
This paper propose a diagnosis architecture that integrates consistency based diagnosis with indu... more This paper propose a diagnosis architecture that integrates consistency based diagnosis with induced time series classi ers, trying to combine the advantages of both methods. Consistency based diagnosis allows fault detection and localization without prior knowledge of the device fault modes. Machine learning techniques are able to induce time series classi ers that may be used to identify fault modes of a dynamic systems. The diagnostician performs fault detection and localization resorting to consistency based diagnosis trough possible conicts. Then, a time series classi er, induced from simulated examples, generates a sequence of faults modes, coherent with the result of the fault localization stage, and ordered by fault modes con dence.
Accurate modeling mechanisms play an important role in model-based diagnosis, and the bond graph ... more Accurate modeling mechanisms play an important role in model-based diagnosis, and the bond graph modeling language has proved to be helpful for this task. In this paper we present an algorithm for automatically derive ARR-like structures, possible conflicts, from the bond graph model of a system. The algorithm uses temporal causal graphs as an intermediate structure to generate the set of possible conflicts. Performance of the algorithm for structural and sensor faults is then studied. Finally, we present another algorithm to automatically derive temporal information in the fault signature matrix for the set of possible conflicts, thus improving the isolation capabilities of the approach.
Complex hybrid systems are present in a large range of engineering applications, like mechanical ... more Complex hybrid systems are present in a large range of engineering applications, like mechanical systems, electrical circuits, or embedded computation systems. The behavior of these systems is made up of continuous and discrete event dynamics that increase the difficulties for accurate and timely online fault diagnosis. The Hybrid Diagnosis Engine (HyDE) offers flexibility to the diagnosis application designer to choose the modeling paradigm and the reasoning algorithms. The HyDE architecture supports the use of multiple modeling paradigms at the component and system level. However, HyDE faces some problems regarding performance in terms of complexity and time. Our focus in this paper is on developing efficient model-based methodologies for online fault diagnosis in complex hybrid systems. To do this, we propose a diagnosis framework where structural model decomposition is integrated within the HyDE diagnosis framework to reduce the computational complexity associated with the fault...
In this chapter, we analyze main problems found by the Artificial Intelligence approach to Model-... more In this chapter, we analyze main problems found by the Artificial Intelligence approach to Model-based diagnosis (DX): the online computation of minimal conflicts by means of an ATMS-like dependency-recording engine, and the need for an extension to deal with dynamic systems diagnosis. To cope with the first problem we will see different options: from extensions to the original GDE to the description of several topological methods, explaining deeply one of them: the Possible Conflict (PC) approach, and its relation with minimal conflicts and ARRs. To cope with the second problem, dynamics, we review the whole set of proposals made to extend Reiter’s formalization and the GDE to dynamic systems: from GDE extensions to the natural extension of topological methods to include temporal information. In this chapter we provide the complete extension of the PCs approach to diagnose dynamic systems, and their relation not only with ARRs, but with another FDI proposals for systems tracking: s...
In this work we introduce DXPCS, a software tool capable of performing consistency-based diagnosi... more In this work we introduce DXPCS, a software tool capable of performing consistency-based diagnosis of continuous dynamic systems whose models can be represented as a set of Ordinary Differential Equations. The diagnosis approach relies upon the Possible Conflict, PC for short, concept. DXPCS is able to automatically build the simulation models for each PC. Single-fault and multiple-fault scenarios, for both parametric and additive faults, can be injected, and studied. DXPCS allows the integration of different algorithms for fault detection, residual generation and evaluation, together with an incremental version of the minimal-hitting set algorithm for fault localization. The software architecture, together with performance results for one simple case study, are provided in this paper.
Reliable and timely fault detection and isolation are necessary tasks to guarantee continuous per... more Reliable and timely fault detection and isolation are necessary tasks to guarantee continuous performance in complex industrial systems, avoiding failure propagation in the system and helping to minimize downtime. Model-based diagnosis fulfils those requirements, and has the additional advantage of using reusable models. However, reusing existing complex non-linear models for diagnosis in large industrial systems is not straightforward. Most of the times the models have been created for other purposes different from diagnosis, and many times the required analytical redundancy is small. In this work we propose to use Possible Conflicts, which is a model decomposition technique, to provide the structure (equations, inputs, outputs, and state variables) of minimal models able to perform fault detection and isolation. Such structural information can be used to design a gray box model by means of state space neural networks. We demonstrate the feasibility of the approach in an evaporator...
Fault Diagnosis of Hybrid Dynamic and Complex Systems
Nowadays hybrid systems are everywhere: vehicles, planes, electronic devices, industrial factorie... more Nowadays hybrid systems are everywhere: vehicles, planes, electronic devices, industrial factories, and so on. All these systems exhibit different behavior patterns depending on the actual operation mode. In this work we propose a framework for fault diagnosis of those dynamic systems characterized by continuous behavior commanded by discrete actuators such as valves, bypasses, relays, etc.
Uploads
Papers