CN118032487A - Digital twin system for on-line monitoring hole extrusion reinforced plate fatigue damage - Google Patents

Abstract

The invention discloses a digital twin system for on-line monitoring hole extrusion reinforced plate fatigue damage, which comprises an initialization unit, an off-line training unit and an on-line deployment unit, wherein the construction of the system specifically comprises the following steps: step 1: planning monitoring points, establishing a numerical model based on material information data and test data of the reinforced plate, and carrying out system initialization operation; step 2: generating a database through a numerical model, and performing offline analysis training by utilizing a neural network; step 3: and installing a dynamic strain gauge on the target plate for real-time monitoring, acquiring real-time local strain information, updating model data in real time, and outputting accumulated damage results to complete on-line deployment and system construction. Through a numerical model of a coupling damage elastoplasticity constitutive equation, a transfer learning neural network, a convolution neural network and a dynamic Bayesian method, fatigue damage of the extruded hole plate can be monitored on line, and self-updating of the system can be realized according to monitoring information so as to ensure accuracy.

Description

Digital twin system for online monitoring of fatigue damage of plate parts strengthened by hole extrusion

Technical Field

The present invention relates to the technical field of hole extrusion fatigue damage monitoring, and in particular to a digital twin system for online monitoring of fatigue damage of hole extrusion reinforced plate parts and a construction method thereof.

Background technique

Hole extrusion is a common structural strengthening process for components with holes. It is widely used in the aviation industry due to its advantages of simple operation, low cost and wide application range. Among the common hole extrusion processes, slotted bushing extrusion is the most widely used due to its advantages of high efficiency and ease of use. The most important slotted bushing extrusion parameters include extrusion amount, initial hole diameter and reaming amount.

Although the hole extrusion process can extend the fatigue life of the hole-containing structure by introducing a residual stress field locally in the hole, the significant stress concentration effect at the hole edge still easily induces fatigue crack initiation, leading to structural failure. Therefore, the current research on the hole extrusion strengthening process for plate parts containing holes needs to be further deepened, especially the need to study how to accurately monitor the accumulated fatigue damage of plate parts containing holes during service, so as to effectively ensure the reliability and integrity of the structure.

After searching, the Chinese patent with application number CN202211740187.X discloses an online monitoring method for the safety and health of high-strength bolts, which mentions a technical solution for monitoring the damage and failure of high-strength bolts;

The Chinese patent application number CN201910921585.3 discloses an acoustic emission detection method for online monitoring of steel corrosion fatigue damage, which mentions a technical solution for monitoring steel corrosion fatigue damage using acoustic emission detection technology;

The Chinese patent application number CN202010927855.4 discloses a multi-scale wind power IGBT reliability assessment method and system taking into account fatigue damage. It proposes a technical means to assess the fatigue damage of wind power converters by extracting the life information of power devices;

However, in terms of online monitoring of fatigue damage, there is no technical solution, model or system for hole extrusion reinforced plates. In addition, the traditional fatigue damage monitoring method cannot perform system self-update based on the real-time sensor information of the monitored object, the accuracy needs to be improved, and the reliability is low.

After searching, the Chinese patent with publication number CN116644559A discloses a method for predicting the service life of steam turbine blades based on a digital twin framework, which proposes a technical means for online monitoring of status information through digital twin technology;

The Chinese patent with publication number CN113569350A discloses a centrifugal compressor impeller fatigue life prediction method based on digital twin, which proposes a technical means for fatigue life prediction through digital twin technology.

However, in terms of digital twin systems, there are currently no technical solutions, models or systems for hole extrusion reinforcement plates. In addition, most of the existing digital twin systems for structural reliability are manually adjusted or cannot adjust the differences between virtual entities and physical entities, involving a heavy empirical component and a low degree of automation.

In addition, in terms of digital twin systems related to fatigue failure, existing systems mainly focus on fatigue crack growth life, pay little attention to fatigue failure processes involving the entire life cycle, and have limited application. In addition, these systems are usually designed only for a single fatigue condition, and cannot consider the diversity of structural processes and external loads, so their scope of application is limited.

Summary of the invention

The purpose of the present invention is to solve the defects in the prior art and to propose a digital twin system for online monitoring of fatigue damage of hole extrusion reinforced plates.

In order to achieve the above object, the present invention adopts the following technical solutions:

According to one aspect of the present invention, a digital twin system for online monitoring of fatigue damage of hole extrusion strengthened plates is provided.

The digital twin system for online monitoring of fatigue damage of hole extrusion reinforcement plate includes an initialization unit, an offline training unit and an online deployment unit, wherein:

The initialization unit includes a fatigue test module, a mechanical model module and a finite element simulation module;

The offline training unit includes a neural network mapping module;

The online deployment unit includes a sensor monitoring module and a Bayesian parameter updating module.

According to another aspect of the present invention, a method for constructing a digital twin system for online monitoring fatigue damage of a hole extrusion reinforced plate is also provided, comprising the following steps:

Step 1: Plan monitoring points, establish a numerical model based on the material information data and test data of the reinforced plate, and perform system initialization operations;

Step 2: Generate a database through the numerical model and use the neural network for offline analysis training;

Step 3: Install dynamic strain gauges on the target panel for real-time monitoring, obtain real-time local strain information, update model data in real time, output cumulative damage results, and complete online deployment and system construction.

Furthermore, the system initialization specifically includes the following steps:

S101: Define fatigue failure criteria based on engineering experience combined with strengthened plate geometry;

S102: Based on the material information of the reinforced plate, static tests are first carried out to obtain the elastic-plastic constitutive parameters of the material, and then standard fatigue tests with stress ratios of -1, 0, and 0.5 are carried out in sequence to calibrate the parameters of the continuous damage mechanics model;

S103: Combining the elastic-plastic constitutive equation with the continuum damage mechanics model to construct the damage elastic-plastic constitutive equation;

S104: compiling the damage elastic-plastic constitutive equation of step S103 into a numerical simulation subroutine;

S105: Conduct fatigue tests on a small number of hole extrusion reinforcement plates, where at least one local strain monitoring point on the reinforcement plate is used to record the test life, infrared thermal imaging during the test, and local strain amplitude monitoring results;

S106: generating a numerical model of the extrusion-strengthened plate corresponding to the test in step S105;

S107: Input the subroutine of step S103 into the numerical model of step S106;

S108: applying the same external load as that in step S105 to the numerical model in step S107, and outputting corresponding fatigue life simulation results and local strain amplitude simulation results for monitoring points;

S109: adjusting the numerical model based on the fatigue life and local strain amplitude test and simulation results generated in step S105 and step S108;

S110: Based on the numerical model adjusted in step S109, further generate numerical models for different hole extrusion processes and external loads, and apply a series of external loads to these models to obtain corresponding fatigue life simulation results and local strain amplitude simulation results for monitoring points;

S111: construct the result of step S110 into a database, marked as database A;

S112: Integrate the infrared thermal imaging image obtained in step S105 and the fatigue damage accumulation simulation result generated in step S108 into a database, marked as database B.

Furthermore, in step 2, the system performs analysis training specifically including the following steps:

S201: using database A to train a transfer learning neural network, the input is the hole extrusion process parameters and the local strain amplitude of the monitoring point, and the output is the fatigue life;

S202: using database B to train a convolutional neural network, the input of which is the infrared thermal imaging image obtained by the test, and the output of which is the fatigue damage accumulation simulation result;

S203: calibrating a phenomenological damage accumulation model using fatigue damage accumulation simulation results;

S204: Setting initial parameters of Gaussian smoothing;

S205: Setting initial parameters of a particle filter based on dynamic Bayesian, wherein a state equation is defined by a damage accumulation model and a monitoring equation is defined by a convolutional neural network.

Furthermore, the steps of online deployment of the system are as follows:

S301: Install dynamic strain gauges to monitor local strain information of target panels in real time;

S302: filtering the strain information obtained in step S301 in real time by Gaussian smoothing;

S303: combining the real-time rain flow counting method to regularize the strain information obtained in step S302 into a regular load spectrum;

S304: inputting the load spectrum of step S303 into the transfer learning network in real time to obtain the fatigue life corresponding to each load block;

S305: inputting a series of fatigue lives obtained in step S304 into a phenomenological damage accumulation model to obtain an estimation result of the accumulated fatigue damage of the target panel;

S306: After the infrared thermal image of the target panel is collected, it is input into the convolutional neural network to obtain an estimated value of the cumulative damage;

S307: Compare the estimated values obtained in step S306 and step S307, thereby defining the weight of each particle in the particle filter;

S308: averaging the high-weight particles in step S307 to update the built-in parameters of the phenomenological damage accumulation model;

S309: Repeat the above steps to estimate the cumulative damage of the target panel in real time. When the estimated value is 1, the system issues an early warning.

Compared with the prior art, the present invention has the following beneficial effects:

The constructed system can interact with the virtual and the real, and update itself through actual monitoring signals, so it can consider the individual factors of the monitored object and make probabilistic predictions;

The constructed system is applicable to a wide range of processes and can be used in principle for all kinds of hole extrusion strengthening processes. It is applicable to a wide range of load conditions and can be deployed under any given process parameters and fatigue load conditions, especially for variable amplitude fatigue loads.

By coupling the numerical model of the damage elastic-plastic constitutive equation (to ensure the adequacy of the neural network training data), the transfer learning neural network and the convolutional neural network (to ensure the feasibility and computational efficiency of the system) and the dynamic Bayesian method (to ensure the system has the ability to self-update), it is possible to monitor the fatigue damage of hole extrusion plates online, and the system can be self-updated according to the monitoring information to ensure accuracy.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

FIG1 is a schematic diagram of an implementation process of constructing a digital twin system for online monitoring fatigue damage of a hole extrusion reinforced plate in an embodiment of the present invention;

2 is a schematic flow chart of a digital twin system for online monitoring of fatigue damage of hole extrusion strengthened plates constructed in an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

As shown in FIG2 , the digital twin system for online monitoring of fatigue damage of hole extrusion strengthened plates includes an initialization unit, an offline training unit, and an online deployment unit, wherein:

The initialization unit includes a fatigue test module, a mechanical model module and a finite element simulation module;

The offline training unit includes a neural network mapping module;

The online deployment unit includes a sensor monitoring module and a Bayesian parameter updating module.

Among them, the initialization unit, the offline training unit and the online deployment unit are connected in sequence.

It needs further explanation:

The initialization unit is used to calibrate the mechanical model (i.e., the damage elastic-plastic constitutive model), and combines with the finite element simulation module to obtain the fatigue failure dynamic data (including the cumulative damage at the dangerous point, the local strain amplitude, and the final failure life) that are difficult to determine by experiments. The sensor monitoring module also records the real-time monitoring signals (infrared thermal imaging signals and local strain signals).

An offline training unit, used to construct a mapping relationship between process parameters and test monitoring signals and cumulative damage and fatigue life using convolutional neural networks and transfer learning neural networks as learners, and save the mapping relationship to a local host;

The online deployment unit is used for the sensing monitoring model to monitor the real-time local strain signals of the panels and periodically monitor the infrared thermal imaging signals of the panels;

The appropriately processed monitoring signals are then imported into the neural network mapping relationship in the local host, and the estimation of fatigue damage of the plate is derived;

At the same time, the built-in parameters of the damage estimation model are updated in real time based on the Bayesian parameter updating module.

In order to better understand the technical solution of the present application, according to an embodiment of the present invention, a method for constructing a digital twin system for online monitoring fatigue damage of hole extrusion reinforced plates is also provided.

1 , a method for constructing a digital twin system for online monitoring of fatigue damage of a hole extrusion reinforced plate comprises the following steps:

Step 1: Plan monitoring points, establish a numerical model based on the material information data and test data of the reinforced plate, and perform system initialization operations;

Step 2: Generate a database through the numerical model and use the neural network for offline analysis training;

Step 3: Install dynamic strain gauges on the target panel for real-time monitoring, obtain real-time local strain information, update model data in real time, output cumulative damage results, and complete online deployment and system construction.

In this embodiment, dynamic strain gauges are installed at the monitoring points to monitor the local strain information of the target panel in real time.

In a specific embodiment of the present application, system initialization specifically includes the following steps:

S101: Define fatigue failure criteria based on engineering experience combined with the geometry of the reinforced plate (e.g. crack extension to 10 mm);

S102: Based on the material information of the reinforced plate, static tests are first carried out to obtain the elastic-plastic constitutive parameters of the material, and then standard fatigue tests with stress ratios of -1, 0, and 0.5 are carried out in sequence to calibrate the parameters of the continuous damage mechanics model;

S103: Combine the elastic-plastic constitutive equation with the continuum damage mechanics model to construct the damage elastic-plastic constitutive equation;

S104: Write the damage elastic-plastic constitutive equation of step S103 as a numerical simulation subroutine

S105: Conduct fatigue tests on a small number of hole extrusion reinforcement plates, where at least one local strain monitoring point is located on the reinforcement plate, and record the test life, infrared thermal imaging during the test, and local strain amplitude monitoring results;

S106: generating a numerical model of the extrusion-strengthened plate corresponding to the test in step S105;

S107: Input the subroutine of step S103 into the numerical model of step S106;

S108: applying the same external load as that in step S105 to the numerical model in step S107, and outputting corresponding fatigue life simulation results and local strain amplitude simulation results for the monitoring points;

S109: adjusting the numerical model based on the fatigue life and local strain amplitude test and simulation results generated in step S105 and step S108;

S110: Based on the numerical model adjusted in step S109, further generate numerical models for different hole extrusion processes and external loads, and apply a series of external loads to these models to obtain corresponding fatigue life simulation results and local strain amplitude simulation results for monitoring points;

S111: construct the result of step S110 into a database, marked as database A;

S112: Integrate the infrared thermal imaging image obtained in step S105 and the fatigue damage accumulation simulation result generated in step S108 into a database, marked as database B.

In a specific embodiment of the present application, in step 2, the system performs analysis training specifically including the following steps:

S201: using database A to train a transfer learning neural network, the input is the hole extrusion process parameters and the local strain amplitude of the monitoring point, and the output is the fatigue life;

S202: using database B to train a convolutional neural network, the input of which is the infrared thermal imaging image obtained by the test, and the output of which is the fatigue damage accumulation simulation result;

S203: calibrating a phenomenological damage accumulation model using fatigue damage accumulation simulation results;

S204: Setting initial parameters of Gaussian smoothing;

S205: Setting initial parameters of a particle filter based on dynamic Bayesian, wherein the state equation is defined by a damage accumulation model and the monitoring equation is defined by a convolutional neural network.

In a specific embodiment of the present application, the steps of online deployment of the system are as follows:

S301: Install dynamic strain gauges to monitor local strain information of the target plate in real time;

S302: filtering the strain information obtained in step S301 in real time by Gaussian smoothing;

S303: combining the real-time rain flow counting method to regularize the strain information obtained in step S302 into a regular load spectrum;

S304: inputting the load spectrum of step S303 into the transfer learning network in real time to obtain the fatigue life corresponding to each load block;

S305: inputting a series of fatigue lives obtained in step S304 into the phenomenological damage accumulation model to obtain an estimation result of the accumulated fatigue damage of the target panel (note that this step is carried out in combination with the particle filter, that is, the phenomenological damage accumulation model must be discretized and written into a damage accumulation form related to fatigue life);

S306: After the infrared thermal image of the target panel is collected, it is input into the convolutional neural network to obtain an estimated value of the cumulative damage;

S307: Compare the estimated values obtained in step S306 and step S307, thereby defining the weight of each particle in the particle filter;

S308: averaging the high-weight particles in step S307 to update the built-in parameters of the phenomenological damage accumulation model;

S309: Repeat the above steps to estimate the cumulative damage of the target panel in real time. When the estimated value is 1, the system issues an early warning.

The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims (5)

1. A digital twin system for online monitoring of fatigue damage of hole extrusion strengthened plates, characterized in that it includes an initialization unit, an offline training unit and an online deployment unit, wherein: The initialization unit includes a fatigue test module, a mechanical model module and a finite element simulation module; The offline training unit includes a neural network mapping module; The online deployment unit includes a sensor monitoring module and a Bayesian parameter updating module. 2. The digital twin system for online monitoring of fatigue damage of hole extrusion reinforcement plate according to claim 1 is characterized in that the construction method of the system specifically comprises the following steps: Step 1: Plan monitoring points, establish a numerical model based on the material information data and test data of the reinforced plate, and perform system initialization operations; Step 2: Generate a database through the numerical model and use the neural network for offline analysis training; Step 3: Install dynamic strain gauges on the target panel for real-time monitoring, obtain real-time local strain information, update model data in real time, output cumulative damage results, and complete online deployment and system construction. 3. The digital twin system for online monitoring of fatigue damage of hole extrusion reinforcement plate according to claim 2 is characterized in that system initialization specifically includes the following steps: S101: Define fatigue failure criteria based on engineering experience combined with enhanced plate geometry; S102: Based on the material information of the reinforced plate, static tests are first carried out to obtain the elastic-plastic constitutive parameters of the material, and then standard fatigue tests with stress ratios of -1, 0, and 0.5 are carried out in sequence to calibrate the parameters of the continuous damage mechanics model; S103: Combining the elastic-plastic constitutive equation with the continuum damage mechanics model to construct the damage elastic-plastic constitutive equation; S104: compiling the damage elastic-plastic constitutive equation of step S103 into a numerical simulation subroutine; S105: Conduct fatigue tests on a small number of hole extrusion reinforcement plates, where at least one local strain monitoring point on the reinforcement plate is used to record the test life, infrared thermal imaging during the test, and local strain amplitude monitoring results; S106: Generate a numerical model of the extrusion-strengthened plate corresponding to the test in step S105; S107: Input the subroutine of step S103 into the numerical model of step S106; S108: applying the same external load as that in step S105 to the numerical model in step S107, and outputting corresponding fatigue life simulation results and local strain amplitude simulation results for the monitoring points; S109: adjusting the numerical model based on the fatigue life and local strain amplitude test and simulation results generated in step S105 and step S108; S110: Based on the numerical model adjusted in step S109, further generate numerical models for different hole extrusion processes and external loads, and apply a series of external loads to these models to obtain corresponding fatigue life simulation results and local strain amplitude simulation results for monitoring points; S111: construct the result of step S110 into a database, marked as database A; S112: Integrate the infrared thermal imaging image obtained in step S105 and the fatigue damage accumulation simulation result generated in step S108 into a database, marked as database B. 4. The digital twin system for online monitoring of fatigue damage of hole extrusion strengthening plates according to claim 3 is characterized in that, in step 2, the system performs analysis training specifically including the following steps: S201: using database A to train a transfer learning neural network, the input is the hole extrusion process parameters and the local strain amplitude of the monitoring point, and the output is the fatigue life; S202: using database B to train a convolutional neural network, with the input being the infrared thermal imaging image obtained by the test, and the output being the fatigue damage accumulation simulation result; S203: calibrating a phenomenological damage accumulation model using fatigue damage accumulation simulation results; S204: Setting initial parameters of Gaussian smoothing; S205: Setting initial parameters of a particle filter based on dynamic Bayesian, wherein a state equation is defined by a damage accumulation model and a monitoring equation is defined by a convolutional neural network. 5. The digital twin system for online monitoring of fatigue damage of hole extrusion reinforcement plates according to claim 4 is characterized in that the steps of online deployment of the system are as follows: S301: Install dynamic strain gauges to monitor local strain information of the target plate in real time; S302: filtering the strain information obtained in step S301 in real time by Gaussian smoothing; S303: combining the real-time rain flow counting method to regularize the strain information obtained in step S302 into a regular load spectrum; S304: inputting the load spectrum of step S303 into the transfer learning network in real time to obtain the fatigue life corresponding to each load block; S305: inputting a series of fatigue lives obtained in step S304 into a phenomenological damage accumulation model to obtain an estimation result of the accumulated fatigue damage of the target panel; S306: After the infrared thermal image of the target panel is collected, it is input into the convolutional neural network to obtain an estimated value of the cumulative damage; S307: Compare the estimated values obtained in step S306 and step S307, thereby defining the weight of each particle in the particle filter; S308: averaging the high-weight particles in step S307 to update the built-in parameters of the phenomenological damage accumulation model; S309: Repeat the above steps to estimate the cumulative damage of the target panel in real time. When the estimated value is 1, the system issues an early warning.