Method and device for evaluating state of film acoustic vibration sensor of micro electro mechanical system

Abstract

The embodiment of the invention provides a method and device for evaluating the state of a film acoustic vibration sensor of a micro electro mechanical system (MEMS), and the method specifically comprises the steps: carrying out the coupling modeling according to the dynamic response of the micro-nano scale data of the MEMS film acoustic vibration sensor in different damage states, and building a damage fingerprint matrix; performing dynamic transfer learning according to historical accelerated aging data and operation data of the MEMS film acoustic vibration sensor, and establishing a digital twinborn evaluation engine; obtaining target micro-nano scale dynamic response data, target field operation data and target working condition data corresponding to the target MEMS film acoustic vibration sensor; and performing decision optimization on the target micro-nano scale data, the target field operation data and the target working condition data through the damage fingerprint matrix, the digital twin evaluation engine and the working condition context to obtain a dynamic health index and an autonomous maintenance instruction set corresponding to the target MEMS film acoustic vibration sensor. Through multi-physics field coupling modeling and a digital twin engine, sensor evaluation is realized.

Description

Technical Field

[0001] The present invention relates to the field of sensor state evaluation, and in particular to a method and device for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system. Background Art

[0002] In recent years, research has focused on the condition assessment of MEMS (Micro-Electro-Mechanical Systems) thin-film acoustic vibration sensors. Condition assessment of MEMS thin-film acoustic vibration sensors is a key technology for ensuring their stable and reliable performance. Leveraging MEMS technology, these sensors integrate a thin-film structure with acoustic vibration sensing capabilities and are widely used in industrial monitoring and environmental testing. The assessment process primarily involves collecting vibration and acoustic signals from the sensor during operation and, through signal processing and machine learning techniques, extracting characteristic parameters to determine its health status. Traditional methods rely on time-domain and frequency-domain analysis to identify anomalies by comparing them against preset thresholds. Machine learning uses historical data to train models and mine data patterns for intelligent diagnosis. Emerging digital twin technology builds virtual models to simulate sensor operation, providing more comprehensive condition information. Condition assessment not only enables timely detection of potential faults but also generates a dynamic health index and enables the development of autonomous maintenance instructions, thereby reducing the risk of equipment downtime and improving system efficiency and safety.

[0003] Some methods, based on traditional signal processing techniques, analyze vibration signals collected by sensors in the time and frequency domains to extract characteristic parameters such as frequency, amplitude, and power spectrum. These methods then use preset thresholds to determine whether the sensor is faulty. For example, if the amplitude of a characteristic frequency in the vibration signal exceeds a preset threshold, the sensor may be considered to be abnormal. Other studies have incorporated machine learning algorithms, using extensive historical data to train models to classify and predict sensor status. These models can learn the characteristic patterns of sensors under different conditions, enabling more accurate assessment of their status.

[0004] However, existing technologies still have many shortcomings in accurately evaluating sensor status, automatically generating results, and intelligent maintenance. Traditional signal processing methods rely on preset thresholds and are difficult to adapt to complex and changeable actual working conditions. Because the normal vibration characteristics of sensors vary greatly under different working environments and load conditions, fixed thresholds can easily lead to misjudgments or missed judgments. Although machine learning-based evaluation methods can learn data patterns, they are highly dependent on data and the evaluation results are not accurate enough. Most existing methods do not fully consider the multi-physical field coupling effects of MEMS thin film acoustic vibration sensors at the micro-nano scale. They only analyze from the perspective of a single physical field or a simple combination, and cannot fully reflect the damage mechanism and state changes in their actual work, resulting in a lack of physical interpretability of the evaluation results. Summary of the Invention

[0005] In view of the above problems, the present application is proposed to provide a method and device for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system, which overcomes the above problems or at least partially solves the above problems, including: A condition assessment of a thin film acoustic vibration sensor of a micro-electromechanical system, comprising: performing coupled modeling based on the dynamic response of micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states, and establishing a damage fingerprint matrix; Dynamically transfer learning is performed based on historical accelerated aging data and operating data of MEMS thin film acoustic vibration sensors to establish a digital twin evaluation engine; Acquire target micro-nanoscale dynamic response data, target field operation data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor; The target micro-nanoscale data, the target field operation data and the target working condition data are subject to decision optimization through the damage fingerprint matrix, the digital twin evaluation engine and the working condition context, so as to obtain a dynamic health index and an autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor.

[0006] Preferably, the step of performing coupled modeling based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix includes: Constructing a multi-physics constitutive relationship based on the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states; Designing a cross-scale coupling interface based on the multi-physics constitutive relationship and the interaction mechanism between different physical fields; Solve and verify the cross-scale model according to the cross-scale coupling interface to build a verified cross-scale physical field coupling model; The characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states are calculated based on the verified cross-scale physical field coupling model, and the damage fingerprint matrix with physical interpretability is constructed; wherein the characteristic parameters include vibration frequency offset, amplitude attenuation rate and electrical performance change.

[0007] Preferably, the multi-physics constitutive relationship includes piezoelectric effect and thermal stress effect, the cross-scale coupling interface includes a piezoelectric effect coupling interface and a thermal stress effect coupling interface, and the step of designing the cross-scale coupling interface based on the multi-physics constitutive relationship and the interaction mechanism between different physical fields includes: A piezoelectric effect coupling interface is designed based on the piezoelectric effect; wherein the piezoelectric effect coupling interface is used to use the strain generated by the mechanical field as the input of the electric field, and simultaneously use the electric field generated by the electric field as the feedback of the mechanical field; A thermal stress effect coupling interface is designed based on the thermal stress effect; wherein the thermal stress effect coupling interface is used to use the thermal stress caused by the temperature change generated by the thermal field as the input of the mechanical field, and at the same time use the stress generated by the mechanical field as the feedback of the thermal field.

[0008] Preferably, the step of performing dynamic migration learning based on historical accelerated aging data and operating data of the MEMS thin film acoustic vibration sensor to establish a digital twin evaluation engine includes: Performing data preprocessing and feature extraction based on the historical accelerated aging data to construct a neural differential equation model; Aligning the operating data with the historical accelerated aging data and adapting them to the domain to obtain aligned operating data; Performing dynamic transfer learning based on the aligned operating data and the neural differential equation model to construct an updated neural differential equation model; A self-evolving digital twin evaluation engine is constructed based on the updated neural differential equation model.

[0009] Preferably, the step of constructing a self-evolving digital twin evaluation engine based on the updated neural differential equation model includes: Establishing a digital twin of a MEMS thin film acoustic vibration sensor, integrating the updated neural differential equation model with the digital twin of the MEMS thin film acoustic vibration sensor, and monitoring and evaluating the state of the MEMS thin film acoustic vibration sensor in real time to obtain an evaluation result; The evaluation results are fed back to the digital twin of the MEMS thin film acoustic vibration sensor to build a digital twin evaluation engine with self-evolution capability.

[0010] Preferably, the step of performing decision optimization on the target micro-nanoscale data, the target field operation data, and the target operating condition data through the damage fingerprint matrix, the digital twin assessment engine, and the operating condition context to obtain a dynamic health index and an autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor includes: According to the damage fingerprint matrix and the working condition context, a feature vector is spliced ​​to construct a state assessment agent, a maintenance decision agent and a resource management agent; Obtaining an optimized multi-agent reinforcement learning model based on the decision optimization of the state assessment agent, the maintenance decision agent, and the resource management agent based on the digital twin evaluation engine; The target micro-nanoscale data, the target on-site operation data, and the target working condition data are input into the optimized multi-agent reinforcement learning model to generate a decision result, and a dynamic health index and an autonomous maintenance instruction set are generated.

[0011] Preferably, the step of inputting the target micro-nanoscale data, the target field operation data, and the target operating condition data into the optimized multi-agent reinforcement learning model to generate a dynamic health index and an autonomous maintenance instruction set includes: Extracting the decision results to obtain the current damage level, remaining service life prediction and performance degradation trend; generating a dynamic health index according to a preset health status quantification rule, the current damage level, the remaining service life prediction, and the performance degradation trend; An autonomous maintenance instruction set is obtained according to the decision result; wherein the autonomous maintenance instruction set includes preventive maintenance, corrective maintenance and predictive maintenance.

[0012] To implement this application, a thin film acoustic vibration sensor state evaluation device of a micro-electromechanical system is also provided, comprising: The damage fingerprint matrix establishment module is used to conduct coupled modeling based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish the damage fingerprint matrix; A digital twin evaluation engine establishment module is used to perform dynamic transfer learning based on historical accelerated aging data and operating data of MEMS thin film acoustic vibration sensors to establish a digital twin evaluation engine; A data acquisition module is used to acquire target micro-nanoscale dynamic response data, target field operation data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor; The dynamic health index and autonomous maintenance instruction set module is used to optimize the decision-making of the target micro-nanoscale data, the target field operation data and the target operating condition data through the damage fingerprint matrix, the digital twin evaluation engine and the operating condition context, and obtain the dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor.

[0013] To implement the present application, a computer electronic device is also included, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, the steps of evaluating the state of the thin film acoustic vibration sensor of the micro-electromechanical system are implemented.

[0014] To implement the present application, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system are implemented.

[0015] This application has the following advantages: In the embodiments of the present application, compared to the problems in the prior art such as the inability to accurately assess sensor status, inaccurate assessment results, and the inability to fully reflect damage mechanisms and state changes in actual operation from the perspective of only a single physical field or a simple combination, the present application provides a multi-physics analysis solution, specifically: coupling modeling based on the dynamic response of the micro- and nano-scale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix; dynamic transfer learning based on the historical accelerated aging data and operating data of the MEMS thin film acoustic vibration sensor to establish a digital twin assessment engine; obtaining target micro- and nano-scale dynamic response data, target field operating data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor; and optimizing the target micro- and nano-scale data, target field operating data, and target operating condition data through the damage fingerprint matrix, the digital twin assessment engine, and the operating condition context to obtain a dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor. The present invention proposes a MEMS thin film acoustic vibration sensor status assessment method that realizes dynamic sensor health status assessment and intelligent maintenance through multi-physics field coupling modeling, digital twin engine construction, and multi-agent decision optimization. Based on the geometric parameters and material properties of the sensor's micro-nanostructure, a multi-physics constitutive relationship is constructed, and a cross-scale coupling interface for piezoelectric and thermal stress effects is designed. The cross-scale physics coupling model is then verified using actual operating environment parameters. Characteristic parameters such as vibration frequency offset are calculated to form a damage fingerprint matrix, addressing the bias issues inherent in traditional single-physics analysis. By integrating historical accelerated aging data with field operation data and using dynamic transfer learning driven by neural differential equations, a self-evolving digital twin evaluation engine is constructed. This engine uses incremental learning to adjust parameters in real time, mitigating the impact of historical data, adapting to performance variations across operating conditions and aging stages, and improving model generalization. The damage fingerprint matrix and operating condition contextual features are concatenated into a feature vector, which is then fed into three types of intelligent agents. The state space, action space, and reward function are defined. Through multi-agent policy gradient optimization, decision-making is optimized to generate a dynamic health index and an autonomous instruction set encompassing preventive, corrective, and predictive maintenance. This achieves closed-loop management from monitoring to maintenance, improving maintenance efficiency and reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for the description of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] Figure 1 This is a flowchart of a method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system provided by an embodiment of the present application; Figure 2 This is a flow chart of a method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system provided in one embodiment of the present application; Figure 3 This is a structural block diagram of a thin film acoustic vibration sensor state evaluation device of a micro-electromechanical system provided by an embodiment of the present application; Figure 4 This is a schematic structural diagram of a computer device provided by one embodiment of the present invention; 1. Computer equipment; 2. External devices; 3. Processing unit; 4. Bus; 5. Network adapter; 6. I / O interface; 7. Display; 8. Memory; 9. Random access memory; 10. Cache memory; 11. Storage system; 12. Programs / utilities; 13. Program modules. DETAILED DESCRIPTION

[0018] To make the objectives, features, and advantages of this application more readily apparent, the present application is further described below in conjunction with the accompanying drawings and specific embodiments. It is apparent that the embodiments described are only a portion of the embodiments of this application, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments in this application without inventive effort are also within the scope of protection of this application.

[0019] By analyzing existing technologies, the inventors discovered that in recent years, there has been some research on the state assessment technology of MEMS thin film acoustic vibration sensors. Some methods are based on traditional signal processing technology. By performing time domain and frequency domain analysis on the vibration signals collected by the sensor, characteristic parameters such as frequency, amplitude, and power spectrum are extracted, and then the sensor is judged to be faulty based on preset thresholds. For example, when the amplitude of a characteristic frequency of the vibration signal exceeds the set threshold, it is determined that the sensor may be abnormal. Some studies have also introduced machine learning algorithms, using large amounts of historical data to train models to classify and predict the state of the sensor. These models can learn the characteristic patterns of the sensor in different states, thereby achieving a more accurate assessment of the sensor state. In addition, digital twin technology has also been gradually applied to the field of sensor state assessment. By constructing a virtual model of the sensor, its behavior and performance in the actual working environment are simulated, providing more comprehensive information support for state assessment.

[0020] However, existing technologies still have many shortcomings in accurately assessing the status of MEMS thin film acoustic vibration sensors and generating dynamic health indices and autonomous maintenance instruction sets to achieve precise status monitoring and intelligent maintenance. Traditional signal processing methods rely on preset thresholds and are difficult to adapt to complex and changeable actual working conditions. Under different working environments and load conditions, the normal vibration characteristics of sensors vary greatly, and fixed thresholds are prone to misjudgment or missed judgments. Although machine learning-based evaluation methods can learn data patterns, they are highly dependent on data. If the training data cannot cover all possible working conditions and damage conditions, the generalization ability of the model will be affected, and the evaluation results will be inaccurate. Moreover, most existing methods do not fully consider the multi-physical field coupling effects at the micro-nano scale of MEMS thin film acoustic vibration sensors. They only analyze from the perspective of a single physical field or a simple combination, and cannot fully reflect the damage mechanism and state changes of the sensor in actual work, resulting in a lack of physical interpretability of the evaluation results.

[0021] In the embodiments of the present application, compared to the problems in the prior art such as the inability to accurately assess sensor status, inaccurate assessment results, and the inability to fully reflect damage mechanisms and state changes in actual operation from the perspective of only a single physical field or a simple combination, the present application provides a multi-physics analysis solution, specifically: coupling modeling based on the dynamic response of the micro- and nano-scale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix; dynamic transfer learning based on the historical accelerated aging data and operating data of the MEMS thin film acoustic vibration sensor to establish a digital twin assessment engine; obtaining target micro- and nano-scale dynamic response data, target field operating data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor; and optimizing the target micro- and nano-scale data, target field operating data, and target operating condition data through the damage fingerprint matrix, the digital twin assessment engine, and the operating condition context to obtain a dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor. The present invention proposes a MEMS thin film acoustic vibration sensor status assessment method that realizes dynamic sensor health status assessment and intelligent maintenance through multi-physics field coupling modeling, digital twin engine construction, and multi-agent decision optimization. Based on the geometric parameters and material properties of the sensor's micro-nanostructure, a multi-physics constitutive relationship is constructed, and a cross-scale coupling interface for piezoelectric and thermal stress effects is designed. The cross-scale physics coupling model is then verified using actual operating environment parameters. Characteristic parameters such as vibration frequency offset are calculated to form a damage fingerprint matrix, addressing the bias issues inherent in traditional single-physics analysis. By integrating historical accelerated aging data with field operation data and using dynamic transfer learning driven by neural differential equations, a self-evolving digital twin evaluation engine is constructed. This engine uses incremental learning to adjust parameters in real time, mitigating the impact of historical data, adapting to performance variations across operating conditions and aging stages, and improving model generalization. The damage fingerprint matrix and operating condition contextual features are concatenated into a feature vector, which is then fed into three types of intelligent agents. The state space, action space, and reward function are defined. Through multi-agent policy gradient optimization, decision-making is optimized to generate a dynamic health index and an autonomous instruction set encompassing preventive, corrective, and predictive maintenance. This achieves closed-loop management from monitoring to maintenance, improving maintenance efficiency and reducing costs.

[0022] Reference Figure 1 , shows a flowchart of the steps of evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system provided by an embodiment of the present application, which specifically includes the following steps: S110, conduct coupled modeling based on the dynamic response of micro-nanoscale data of MEMS thin film acoustic vibration sensors under different damage states, and establish a damage fingerprint matrix; S120: Dynamically transfer learning based on historical accelerated aging data and operating data of MEMS thin film acoustic vibration sensors to establish a digital twin evaluation engine; S130, acquiring target micro-nanoscale dynamic response data, target field operation data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor; S140, performing decision optimization on the target micro-nanoscale data, the target field operation data, and the target operating condition data using the damage fingerprint matrix, the digital twin assessment engine, and the operating condition context to obtain a dynamic health index and an autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor; Next, the state evaluation of the thin film acoustic vibration sensor of the micro-electromechanical system in this exemplary embodiment will be further described.

[0023] In one embodiment of the present invention, the specific process of "performing coupled modeling based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix" in step S110 can be further explained in combination with the following description.

[0024] As an example, based on the micro- and nanoscale dynamic response of MEMS thin film acoustic vibration sensors, cross-scale physical field coupling modeling is performed to obtain a damage fingerprint matrix with physical interpretability.

[0025] Step S110 specifically includes the following steps: Constructing a multi-physics constitutive relationship based on the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states; Design cross-scale coupling interfaces based on the multi-physics constitutive relations and the interaction mechanisms between different physical fields. Solve and verify the cross-scale model according to the cross-scale coupling interface to build a verified cross-scale physical field coupling model; The characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states are calculated based on the verified cross-scale physical field coupling model, and the damage fingerprint matrix with physical interpretability is constructed; wherein the characteristic parameters include vibration frequency offset, amplitude attenuation rate and electrical performance change.

[0026] As an example, a multi-physics field constitutive relationship is constructed based on the geometric parameters and material properties of the micro-nano structure of the MEMS thin film acoustic vibration sensor; a cross-scale coupling interface is designed based on the multi-physics field constitutive relationship according to the interaction mechanism between different physical fields; according to the actual working environment and boundary conditions of the sensor, a cross-scale model is solved and verified to obtain a verified cross-scale physical field coupling model, which is used to simulate the dynamic response of the MEMS thin film acoustic vibration sensor in the actual working environment; based on the verified cross-scale physical field coupling model, the characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states are calculated and combined into a damage fingerprint matrix with physical interpretability; the characteristic parameters include vibration frequency offset, amplitude attenuation rate and electrical performance change.

[0027] As an example, the present invention fully considers the interaction of multiple physical fields at the micro-nano scale of MEMS thin film acoustic vibration sensors through cross-scale physical field coupling modeling, more accurately simulates the dynamic response of MEMS thin film acoustic vibration sensors in actual working environments, and provides a reliable basis for subsequent state assessment. The present invention also constructs a damage fingerprint matrix by analyzing the characteristic parameters under different damage states, which can reflect the damage mechanism of MEMS thin film acoustic vibration sensors under the action of different physical fields, and provide a physical explanation for understanding the damage evolution process of MEMS thin film acoustic vibration sensors. Among them, the damage fingerprint matrix with physical interpretability can more accurately characterize the damage state of the sensor, and greatly improves the accuracy of state assessment compared to traditional methods based only on a single physical field or simple features.

[0028] For example, during operation, the micro-nanostructure of a MEMS thin-film acoustic vibration sensor is subject to the combined effects of multiple physical fields, including mechanical, electrical, and thermal fields. Complex interaction mechanisms exist between these different physical fields, such as the piezoelectric effect (the coupling of mechanical and electrical fields) and the thermal stress effect (the coupling of mechanical and thermal fields). These couplings can affect sensor performance and damage evolution.

[0029] In one embodiment of the present invention, the specific process of "constructing a multi-physics field constitutive relationship based on the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states" can be further explained in combination with the following description.

[0030] As described in the following steps, the micro-nanostructure geometric parameters and material characteristic parameters of the MEMS thin film acoustic vibration sensor are collected; the micro-nanostructure geometric parameters include the thickness, length, width and shape of the MEMS film; the material characteristic parameters include the elastic modulus, Poisson's ratio, density and piezoelectric coefficient of the MEMS film; based on the micro-nanostructure geometric parameters and material characteristic parameters, combined with basic physical theories, the elasticity constitutive equation, Maxwell's equations and heat conduction equation are established respectively; the elasticity constitutive equation is used to describe the relationship between stress and strain; the Maxwell equations are used to describe the relationship between the electric field and the magnetic field; the heat conduction equation is used to describe the relationship between temperature and heat flow; based on the elasticity constitutive equation, Maxwell's equations and heat conduction equation, a constitutive relationship model is constructed to describe the multi-physics field constitutive relationship of the micro-nanostructure of the MEMS thin film acoustic vibration sensor.

[0031] As an example, the first step includes collecting the micro-nanostructure geometric parameters and material characteristic parameters of the MEMS thin film acoustic vibration sensor. In this embodiment, the micro-nanostructure geometric parameters include the thickness, length, width, and shape of the MEMS film, and the material characteristic parameters include the elastic modulus, Poisson's ratio, density, and piezoelectric coefficient of the MEMS film.

[0032] As an example, the second step includes: based on the geometric parameters of the micro-nano structure and the material characteristic parameters, combined with basic physical theories, the elastic mechanics constitutive equation, Maxwell's equations and heat conduction equation are established respectively.

[0033] In this embodiment, the constitutive equation of elasticity is used to describe the relationship between stress and strain, the Maxwell equations are used to describe the relationship between the electric field and the magnetic field, and the heat conduction equation is used to describe the relationship between temperature and heat flow.

[0034] As an example, the third step includes: constructing a constitutive relationship model based on the elastic mechanics constitutive equation, Maxwell's equations and heat conduction equation to describe the multi-physics field constitutive relationship of the micro-nano structure of the MEMS thin film acoustic vibration sensor.

[0035] To accurately simulate the dynamic response of a MEMS thin film acoustic vibration sensor in an actual operating environment, the present invention constructs a cross-scale physical field coupling model. First, the geometric parameters of the micro-nanostructure of the MEMS thin film acoustic vibration sensor, such as the thickness, length, width, and shape of the film, as well as material properties such as elastic modulus, Poisson's ratio, density, and piezoelectric coefficient, are collected. Based on these parameters and combined with basic physical theories, the elastic mechanics constitutive equation is established to describe the relationship between stress and strain, the Maxwell equations are used to describe the relationship between electric field and magnetic field, and the heat conduction equation is used to describe the relationship between temperature and heat flow. A constitutive relationship model is then constructed to describe the multi-physics field constitutive relationship of the sensor's micro-nanostructure.

[0036] In one embodiment of the present invention, the specific process of “designing a cross-scale coupling interface based on the multi-physics constitutive relationship and the interaction mechanism between different physical fields” can be further explained in combination with the following description.

[0037] As described in the following steps, the multi-physics field constitutive relationship includes piezoelectric effect and thermal stress effect, and the cross-scale coupling interface includes a piezoelectric effect coupling interface and a thermal stress effect coupling interface; a piezoelectric effect coupling interface is designed based on the piezoelectric effect; wherein the piezoelectric effect coupling interface is used to use the strain generated by the mechanical field as the input of the electrical field, and simultaneously use the electric field generated by the electrical field as the feedback of the mechanical field; a thermal stress effect coupling interface is designed based on the thermal stress effect; wherein the thermal stress effect coupling interface is used to use the thermal stress caused by the temperature change generated by the thermal field as the input of the mechanical field, and simultaneously use the stress generated by the mechanical field as the feedback of the thermal field.

[0038] As an example, a cross-scale coupling interface is designed based on the piezoelectric effect and thermal stress effect in the MEMS thin film acoustic vibration sensor to realize data transmission and information interaction between different physical fields; the cross-scale coupling interface includes a piezoelectric effect coupling interface and a thermal stress effect coupling interface; in the piezoelectric effect coupling interface, the strain generated by the mechanical field is used as the input of the electrical field, and the electric field generated by the electric field is used as the feedback of the mechanical field; in the thermal stress effect coupling interface, the thermal stress caused by the temperature change generated by the thermal field is used as the input of the mechanical field, and the stress generated by the mechanical field is used as the feedback of the thermal field.

[0039] As an example, it includes: designing a cross-scale coupling interface based on the piezoelectric effect and thermal stress effect in MEMS thin film acoustic vibration sensors to realize data transmission and information interaction between different physical fields.

[0040] In this embodiment, the cross-scale coupling interface includes a piezoelectric effect coupling interface and a thermal stress effect coupling interface. In the piezoelectric effect coupling interface, the strain generated by the mechanical field is used as the input of the electrical field, while the electric field generated by the electrical field is used as the feedback of the mechanical field. In the thermal stress effect coupling interface, the thermal stress caused by temperature changes in the thermal field is used as the input of the mechanical field, while the stress generated by the mechanical field is used as the feedback of the thermal field.

[0041] Based on the interaction mechanisms between different physical fields, this paper designs cross-scale coupling interfaces to enable data transfer and information exchange between different physical fields. For example, in a piezoelectric effect coupling interface, the strain generated by the mechanical field is used as the input of the electrical field, while the electric field generated by the electrical field is used as the feedback of the mechanical field. In a thermal stress effect coupling interface, the thermal stress caused by temperature changes in the thermal field is used as the input of the mechanical field, while the stress generated by the mechanical field is used as the feedback of the thermal field.

[0042] In one embodiment of the present invention, the specific process of "solving and verifying a cross-scale model according to the cross-scale coupling interface to construct a verified cross-scale physical field coupling model" can be further explained in combination with the following description.

[0043] As an example, based on the actual working environment and boundary conditions of the sensor, a cross-scale model is solved and verified to obtain a verified cross-scale physical field coupling model, which is used to simulate the dynamic response of the MEMS thin film acoustic vibration sensor in the actual working environment.

[0044] As described in the following steps, the actual working environment parameters of the MEMS thin film acoustic vibration sensor are input into the constructed cross-scale physical field coupling model. The actual working environment parameters include temperature, pressure, humidity, and boundary conditions. The cross-scale physical field coupling model is solved using the finite element method. The dynamic response of the MEMS thin film acoustic vibration sensor in the actual working environment is tested, and the test results are compared and verified with the solution results. The solution results with an error between the test results and the solution results less than a preset threshold are retained to construct a verified cross-scale physical field coupling model.

[0045] As an example, the first step includes: using the actual working environment parameters of the MEMS thin film acoustic vibration sensor to input into the constructed cross-scale physical field coupling model, the actual working environment parameters including temperature, pressure, humidity and boundary conditions.

[0046] As an example, the second step includes: using the finite element method to solve the cross-scale physical field coupling model, and by testing the dynamic response of the MEMS thin film acoustic vibration sensor in the actual working environment, comparing and verifying the test results with the solution results, retaining the solution results whose error between the test results and the solution results is less than a preset threshold, and obtaining a verified cross-scale physical field coupling model.

[0047] The present invention uses the actual working environment parameters of the MEMS thin film acoustic vibration sensor, such as temperature, pressure, humidity and boundary conditions, and inputs them into a constructed cross-scale physical field coupling model. The model is solved using the finite element method, and the dynamic response of the sensor in the actual working environment is tested. The test results are compared and verified with the solution results, and the solution results with errors less than a preset threshold are retained to construct a verified cross-scale physical field coupling model.

[0048] In one embodiment of the present invention, the specific process of "calculating the characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states according to the verified cross-scale physical field coupling model, and constructing the damage fingerprint matrix with physical interpretability; wherein the characteristic parameters include vibration frequency offset, amplitude attenuation rate and electrical performance change" can be further explained in combination with the following description.

[0049] As an example, based on a verified cross-scale physical field coupling model, the characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states are calculated and combined into a damage fingerprint matrix with physical interpretability.

[0050] As described in the following steps, based on a validated cross-scale physical field coupling model, the characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states are calculated and combined into a physically interpretable damage fingerprint matrix; the characteristic parameters include vibration frequency offset, amplitude attenuation rate, and electrical performance change.

[0051] As an example, in this embodiment, the characteristic parameters include vibration frequency offset, amplitude attenuation rate, and electrical performance change.

[0052] Based on a verified cross-scale physical field coupling model, the present invention calculates the characteristic parameters of MEMS thin film acoustic vibration sensors under different damage states, such as vibration frequency offset, amplitude attenuation rate, and electrical performance change, and combines them into a damage fingerprint matrix with physical interpretability.

[0053] In one specific embodiment, MEMS thin-film acoustic vibration sensors, operating at the micro- and nanoscale, experience complex interactions among multiple physical fields, making the construction of coupled models crucial for damage analysis. The following details the research process, from constitutive relationship construction to damage fingerprint matrix generation. The micro- and nanostructures of MEMS thin-film acoustic vibration sensors involve multiple physical fields, including mechanics and electricity, and their performance is influenced by geometric parameters and material properties. Based on the specific geometric parameters of the sensor's micro- and nanostructure, such as size and shape, and combined with material properties such as elastic modulus, Poisson's ratio, and conductivity, constitutive equations describing the characteristics and interrelationships of various physical fields are established through theoretical derivation and experimental data fitting. These equations lay the theoretical foundation for subsequent research and reveal the internal laws governing these physical fields. The close interactions between different physical fields at the micro- and nanoscale require in-depth analysis of mechanisms such as energy transfer and variable coupling between these fields based on the established multi-field constitutive relationships. Appropriate mathematical methods and computational techniques, such as the finite element method and molecular dynamics simulation, are employed to connect equations across different scales and physical fields, achieving cross-scale coupling and ensuring accurate data and information transfer between these fields. Based on the actual operating environment of the sensor and combined with boundary conditions, numerical calculation methods are used to solve the cross-scale coupling model. The model's accuracy is evaluated by comparing simulation results with actual experimental data, such as the sensor's output signal and vibration response under different operating conditions. Based on the verified cross-scale physical field coupling model, the sensor's operating state under different damage states, such as film cracks and material defects, is simulated. The vibration frequency offset is calculated to reflect changes in structural stiffness; the amplitude attenuation rate reflects energy loss; and the change in electrical performance indicates changes in electrical signal characteristics. These characteristic parameters are combined into a damage fingerprint matrix. Each matrix element has a clear physical meaning, which can intuitively reflect the impact of different damages on sensor performance and provide a quantitative basis for sensor damage detection and health monitoring. The cross-scale physical field coupling model and damage fingerprint matrix constructed through the above steps provide an effective means for damage analysis of MEMS thin film acoustic vibration sensors, help improve the reliability and service life of the sensors, and promote their widespread application in acoustic monitoring, environmental testing and other fields.

[0054] In one embodiment of the present invention, the specific process of "performing dynamic migration learning based on historical accelerated aging data and operation data of the MEMS thin film acoustic vibration sensor to establish a digital twin evaluation engine" in step S120 can be further explained in combination with the following description.

[0055] Step S120 includes the following steps: Performing data preprocessing and feature extraction based on the historical accelerated aging data to construct a neural differential equation model; Aligning the operating data with the historical accelerated aging data and adapting them to the domain to obtain aligned operating data; Performing dynamic transfer learning based on the aligned operating data and the neural differential equation model to construct an updated neural differential equation model; A self-evolving digital twin evaluation engine is constructed based on the updated neural differential equation model.

[0056] As an example, data preprocessing and feature engineering are performed based on historical accelerated aging data to extract historical data features and physical laws, and a neural differential equation model is constructed to capture the dynamic laws of sensor aging; the field operation data and historical accelerated aging data are aligned and domain adapted to obtain aligned field operation data; the aligned field operation data is gradually input into the trained neural differential equation model for dynamic transfer learning to obtain a neural differential equation model that is continuously updated after dynamic transfer learning, which is used to adjust its own parameters in real time according to the field operation data; based on the updated neural differential equation model, a self-evolving digital twin evaluation engine is constructed.

[0057] In a specific embodiment, dynamic transfer learning driven by neural differential equations is performed based on historical accelerated aging data and field operation data to obtain a self-evolving digital twin evaluation engine. The present invention integrates historical accelerated aging data and field operation data through dynamic transfer learning, fully utilizing the advantages of both to improve the generalization ability and adaptability of the model. The present invention also continuously updates the neural differential equation model so that the digital twin evaluation engine can adapt to the performance changes of the sensor in different working conditions and aging stages in real time, ensuring the accuracy and timeliness of the evaluation results. At the same time, the feedback mechanism enables the digital twin evaluation engine to continuously optimize itself according to the evaluation results, has self-evolution capabilities, and can stably evaluate the sensor status over a long period of time.

[0058] As an example, historical accelerated aging data is data acquired through accelerated sensor aging under laboratory conditions. It can reflect the performance changes of sensors at different aging stages. Field operation data is data collected by sensors in actual operating environments and contains various information about the sensors under actual operating conditions.

[0059] In one embodiment of the present invention, the specific process of “performing data preprocessing and feature extraction based on the historical accelerated aging data to construct a neural differential equation model” can be further explained in combination with the following description.

[0060] As described in the following steps, data preprocessing and feature engineering are performed based on historical accelerated aging data to extract historical data features and physical laws, and a neural differential equation model is constructed to capture the dynamic laws of sensor aging.

[0061] This paper preprocesses and features historical accelerated aging data, extracting historical data features and physical laws to construct a neural differential equation model. The neural differential equation combines the advantages of neural networks and differential equations to capture the dynamic patterns of sensor aging.

[0062] In one embodiment of the present invention, the specific process of "aligning the operating data with the historical accelerated aging data and adapting them to the domain to obtain aligned operating data" can be further explained in combination with the following description.

[0063] As described in the following steps, the field operation data is aligned and domain adapted with the historical accelerated aging data to obtain aligned field operation data.

[0064] This method aligns and domain-adapts field data with historical accelerated aging data to produce aligned field data. Since historical accelerated aging data and field data may have distributional differences, domain adaptation can mitigate these differences, enabling the neural differential equation model to better adapt to field data.

[0065] In one embodiment of the present invention, the specific process of "constructing an updated neural differential equation model by performing dynamic transfer learning based on the aligned operating data and the neural differential equation model" can be further explained in combination with the following description.

[0066] As described in the following steps, the aligned field operation data is gradually input into the trained neural differential equation model to perform dynamic transfer learning, thereby obtaining a neural differential equation model that is continuously updated after dynamic transfer learning, including: using an incremental learning method to gradually input the aligned field operation data into the trained neural differential equation model; as the field operation data continues to accumulate, gradually reducing the impact of historical accelerated aging data on the neural differential equation model, increasing the weight of the field operation data, and simultaneously using the mean square error to perform real-time evaluation of the updated neural differential equation model, thereby obtaining a neural differential equation model that is continuously updated after dynamic transfer learning.

[0067] As an example, the aligned field operation data is gradually input into the trained neural differential equation model for dynamic transfer learning, and a neural differential equation model that is continuously updated after dynamic transfer learning is obtained, which is used to adjust its own parameters in real time according to the field operation data.

[0068] As an example, the first step includes: using an incremental learning method to gradually input the aligned field operation data into the trained neural differential equation model.

[0069] As an example, the second step includes: as the field operation data continues to accumulate, the impact of historical accelerated aging data on the neural differential equation model is gradually reduced, the weight of the field operation data is increased, and the mean square error is used to perform real-time evaluation of the updated neural differential equation model to obtain a neural differential equation model that is continuously updated after dynamic transfer learning.

[0070] This method uses an incremental learning approach to gradually input aligned field operation data into a trained neural differential equation model. As field operation data accumulates, the impact of historical accelerated aging data on the model is gradually reduced, while the weight of field operation data is increased. Simultaneously, the updated neural differential equation model is evaluated in real time using mean squared error to ensure its performance. Through dynamic transfer learning, a continuously updated neural differential equation model is obtained.

[0071] In one embodiment of the present invention, the specific process of "constructing a self-evolving digital twin evaluation engine based on the updated neural differential equation model" can be further explained in combination with the following description.

[0072] As described in the following steps, a digital twin of a MEMS thin film acoustic vibration sensor is established, the updated neural differential equation model is integrated with the digital twin of the MEMS thin film acoustic vibration sensor, and the state of the MEMS thin film acoustic vibration sensor is monitored and evaluated in real time to obtain an evaluation result; the evaluation result is fed back to the digital twin of the MEMS thin film acoustic vibration sensor to construct a digital twin evaluation engine with self-evolution capability.

[0073] It should be noted that a digital twin refers to a virtual model created in a virtual space using digital technology that corresponds exactly to a real-world physical object. This model is an intelligent system that can reflect and predict the behavior and state of a physical object under different environments and conditions. The core of a digital twin lies in its ability to reflect changes in the real world in real time and provide decision support through analysis and prediction. A digital twin consists of three main components: Physical entity: refers to physical objects in the real world, such as machinery, equipment, buildings, vehicles, etc.

[0074] Virtual model: It is a digital representation of a physical entity in a virtual space, usually generated through 3D modeling technology and combined with data analysis and simulation technology to simulate the behavior and characteristics of the physical entity.

[0075] Data connection: It is the bridge between physical entities and virtual models. It realizes real-time data collection and transmission through sensors, Internet of Things (IoT) and other technologies, ensuring that the virtual model can reflect the status and behavior of the physical entity in real time.

[0076] Through the combination of the above three parts, the digital twin can achieve accurate simulation and real-time monitoring of physical entities in virtual space.

[0077] As an example, a digital twin of a MEMS thin film acoustic vibration sensor is established, and the updated neural differential equation model is integrated with the digital twin of the MEMS thin film acoustic vibration sensor to monitor and evaluate the status of the MEMS thin film acoustic vibration sensor in real time to obtain an evaluation result; a feedback mechanism is established to feed the evaluation result back to the digital twin of the MEMS thin film acoustic vibration sensor, optimize the digital twin and its parameters, and construct a digital twin evaluation engine with self-evolution capability; the digital twin evaluation engine is used to continuously learn and optimize based on historical accelerated aging data and field operation data to evaluate the status of the MEMS thin film acoustic vibration sensor.

[0078] As an example, a self-evolving digital twin evaluation engine was constructed based on the updated neural differential equation model. The updated neural differential equation model was integrated with the digital twin of a MEMS thin-film acoustic vibration sensor to monitor and evaluate the sensor's status in real time, generating evaluation results. A feedback mechanism was established to feed the evaluation results back to the digital twin of the MEMS thin-film acoustic vibration sensor, optimizing the digital twin and its parameters, ultimately building a self-evolving digital twin evaluation engine.

[0079] In this embodiment, the digital twin evaluation engine is used to continuously learn and optimize based on historical accelerated aging data and field operation data to evaluate the status of the MEMS thin film acoustic vibration sensor.

[0080] In one specific embodiment, the present invention integrates an updated neural differential equation model with a MEMS thin-film acoustic vibration sensor digital twin to monitor and evaluate the sensor's status in real time, generating evaluation results. A feedback mechanism is then established to feed the evaluation results back to the digital twin, optimizing the digital twin and its parameters, thereby constructing a self-evolving digital twin evaluation engine.

[0081] In one embodiment of the present invention, the specific process of "obtaining target micro-nanoscale dynamic response data, target field operation data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor" in step S130 can be further explained in combination with the following description.

[0082] As described in the following steps, micro-nanoscale dynamic response data of the target MEMS thin film acoustic vibration sensor, such as vibration displacement, stress distribution, field operation data, and working condition data, are obtained to provide multi-dimensional data support for subsequent damage analysis.

[0083] In one embodiment of the present invention, the specific process of step S140 of "optimizing the decision on the target micro-nanoscale data, the target field operation data and the target working condition data through the damage fingerprint matrix, the digital twin evaluation engine and the working condition context to obtain the dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor" can be further explained in combination with the following description.

[0084] Step S140 includes the following steps: According to the damage fingerprint matrix and the working condition context, a feature vector is spliced ​​to construct a state assessment agent, a maintenance decision agent and a resource management agent; Obtaining an optimized multi-agent reinforcement learning model based on the decision optimization of the state assessment agent, the maintenance decision agent, and the resource management agent based on the digital twin evaluation engine; The target micro-nanoscale data, the target on-site operation data, and the target working condition data are input into the optimized multi-agent reinforcement learning model to generate a decision result, and a dynamic health index and an autonomous maintenance instruction set are generated.

[0085] As an example, the standardized damage fingerprint matrix is ​​fused with the preprocessed working condition context features by feature splicing, and the eigenvectors are spliced ​​to describe the damage status and current working condition information of the MEMS thin film acoustic vibration sensor; a state assessment agent, a maintenance decision agent and a resource management agent are set; the state assessment agent is used to make a preliminary assessment of the health status of the MEMS thin film acoustic vibration sensor based on the eigenvector using the digital twin assessment engine to obtain the state assessment result; the maintenance decision agent is used to make a preliminary maintenance decision based on the state assessment result and the working condition context in combination with the preset maintenance strategy library; the resource management agent is used to coordinate the resources required for maintenance according to the preliminary maintenance decision; the state space and action space of the state assessment agent, the maintenance decision agent and the resource management agent are defined and reward function; the state space is constructed based on the eigenvector; the action space contains preliminary maintenance decisions; the reward function gives positive rewards when the health status of the MEMS thin film acoustic vibration sensor is improved and the maintenance cost is lower than the preset threshold, and negative rewards otherwise; the behavior of the agent under different working conditions and damage conditions is simulated, the state, action and reward data of the state assessment agent, maintenance decision agent and resource management agent are collected, and the multi-agent reinforcement learning decision optimization is performed through the multi-agent policy gradient to obtain the optimized multi-agent reinforcement learning model, which is used to make the optimal decision based on the damage fingerprint matrix and working condition context of the MEMS thin film acoustic vibration sensor with the support of the digital twin evaluation engine; based on the decision results of the optimized multi-agent reinforcement learning model, a dynamic health index and autonomous maintenance instruction set are generated.

[0086] In a specific embodiment, according to the damage fingerprint matrix and the working condition context, a multi-agent reinforcement learning decision optimization is performed based on the digital twin evaluation engine to obtain a dynamic health index and an autonomous maintenance instruction set. The present invention can comprehensively consider multiple factors such as the damage status of the sensor, working condition information, and maintenance resources through multi-agent reinforcement learning decision optimization to make more reasonable decisions. Among them, the dynamic health index can intuitively reflect the health status of the sensor, provide a convenient monitoring method for operation and maintenance personnel, and facilitate timely understanding of the operating status of the sensor. Among them, the autonomous maintenance instruction set details different types of maintenance operations, realizes the closed-loop management of the entire process from status monitoring to intelligent maintenance, improves maintenance efficiency, reduces maintenance costs, and enhances the operational stability and reliability of the system.

[0087] As an example, according to the damage fingerprint matrix and working condition context, multi-agent reinforcement learning decision optimization is performed based on the digital twin evaluation engine to obtain the dynamic health index and autonomous maintenance instruction set, such as Figure 2 The damage fingerprint matrix contains the characteristic parameters of the sensor under different damage states, which can reflect the damage degree of the sensor. The working condition context contains relevant information about the actual working environment of the sensor, such as temperature, pressure, humidity, etc.

[0088] In one embodiment of the present invention, the specific process of "constructing a state assessment agent, a maintenance decision agent and a resource management agent based on the damage fingerprint matrix and the working condition context being spliced ​​into a feature vector" can be further explained in combination with the following description.

[0089] As described in the following steps, the standardized damage fingerprint matrix is ​​fused with the preprocessed operating condition context features using a feature splicing method to form a feature vector, which is used to describe the damage status and current operating condition information of the MEMS thin film acoustic vibration sensor. The present invention uses a feature splicing method to fuse the standardized damage fingerprint matrix with the preprocessed operating condition context features to form a feature vector, which is used to describe the damage status and current operating condition information of the sensor.

[0090] As shown in the following steps, set up the condition assessment agent, maintenance decision agent, and resource management agent.

[0091] In this embodiment, the state assessment agent is used to perform a preliminary assessment of the health status of the MEMS thin film acoustic vibration sensor based on the feature vector using the digital twin assessment engine to obtain the state assessment results; the maintenance decision agent is used to make preliminary maintenance decisions based on the state assessment results and the working condition context, combined with the preset maintenance strategy library; the resource management agent is used to coordinate the resources required for maintenance based on the preliminary maintenance decisions.

[0092] As described in the following steps, define the state space, action space, and reward function of the state assessment agent, maintenance decision agent, and resource management agent.

[0093] In this embodiment, the state space is constructed based on the feature vector; the action space includes preliminary maintenance decisions; and the reward function provides positive rewards when the health status of the MEMS thin film acoustic vibration sensor is improved and the maintenance cost is lower than a preset threshold, and negative rewards otherwise.

[0094] This paper defines the state space, action space, and reward function for the condition assessment agent, maintenance decision agent, and resource management agent. The state space is constructed based on the feature vector, the action space contains preliminary maintenance decisions, and the reward function provides positive rewards when the sensor health status improves and the maintenance cost falls below a preset threshold, and negative rewards when the maintenance cost falls below a preset threshold.

[0095] In one embodiment of the present invention, the specific process of "obtaining an optimized multi-agent reinforcement learning model based on decision optimization of the status assessment agent, the maintenance decision agent and the resource management agent based on the digital twin assessment engine" can be further explained in combination with the following description.

[0096] As described in the following steps, the behavior of the agent under different working conditions and damage situations is simulated, and the status, action, and reward data of the state assessment agent, maintenance decision agent, and resource management agent are collected. Multi-agent reinforcement learning decision optimization is performed through multi-agent policy gradient. The optimized multi-agent reinforcement learning model is obtained and used to make the optimal decision based on the damage fingerprint matrix and working condition context of the MEMS thin film acoustic vibration sensor with the support of the digital twin evaluation engine.

[0097] The present invention simulates the behavior of intelligent agents under different working conditions and damage situations, collects the status, action and reward data of the state assessment agent, maintenance decision agent and resource management agent, and optimizes the multi-agent reinforcement learning decision through multi-agent policy gradient to obtain an optimized multi-agent reinforcement learning model.

[0098] In one embodiment of the present invention, the specific process of "inputting the target micro-nanoscale data, the target field operation data, and the target working condition data into the optimized multi-agent reinforcement learning model to generate a dynamic health index and an autonomous maintenance instruction set" can be further explained in combination with the following description.

[0099] As described in the following steps, the decision results are extracted to obtain the current damage level, remaining service life prediction and performance degradation trend; a dynamic health index is generated based on preset health status quantification rules, the current damage level, the remaining service life prediction and the performance degradation trend; based on the decision results, an autonomous maintenance instruction set is obtained; wherein the autonomous maintenance instruction set includes preventive maintenance, corrective maintenance and predictive maintenance.

[0100] As an example, the current damage degree, remaining service life prediction and performance degradation trend of the MEMS thin film acoustic vibration sensor are extracted from the decision results of the optimized multi-agent reinforcement learning model; according to the preset health status quantification rules, based on the current damage degree, remaining service life prediction and performance degradation trend, the health status of the MEMS thin film acoustic vibration sensor is mapped to a numerical range to generate a dynamic health index, which is used to intuitively reflect the current health status of the sensor; according to the decision results of the optimized multi-agent reinforcement learning model, the initial maintenance decision is divided into preventive maintenance, corrective maintenance and predictive maintenance, and an autonomous maintenance instruction set is constructed; preventive maintenance includes regular calibration operations, cleaning operations and lubrication operations; corrective maintenance includes replacing faulty parts and repairing damaged structures; predictive maintenance includes arranging maintenance time and content in advance according to the performance degradation trend of the sensor.

[0101] Based on the decision results of the optimized multi-agent reinforcement learning model, a dynamic health index and autonomous maintenance instruction set are generated.

[0102] As an example, the first step includes extracting the current damage degree, remaining service life prediction and performance degradation trend of the MEMS thin film acoustic vibration sensor from the decision results of the optimized multi-agent reinforcement learning model.

[0103] As an example, the second step includes: according to the preset health status quantification rules, based on the current damage level, remaining service life prediction and performance degradation trend, the health status of the MEMS thin film acoustic vibration sensor is mapped to a numerical range to generate a dynamic health index to intuitively reflect the current health status of the sensor.

[0104] As an example, the third step includes: according to the decision results of the optimized multi-agent reinforcement learning model, the initial maintenance decision is divided into preventive maintenance, corrective maintenance and predictive maintenance, and an autonomous maintenance instruction set is constructed.

[0105] In this embodiment, preventive maintenance includes regular calibration operations, cleaning operations, and lubrication operations; corrective maintenance includes replacing faulty parts and repairing damaged structures; and predictive maintenance includes scheduling maintenance time and content in advance based on the performance degradation trend of the sensor.

[0106] In one specific embodiment, the present invention extracts the current sensor damage level, remaining service life prediction, and performance degradation trend from the decision results of an optimized multi-agent reinforcement learning model. Based on preset health status quantification rules, the sensor's health status is mapped to a numerical range, generating a dynamic health index that intuitively reflects the sensor's current health status. Based on the decision results of the optimized multi-agent reinforcement learning model, initial maintenance decisions are classified into preventive maintenance, corrective maintenance, and predictive maintenance, and an autonomous maintenance instruction set is constructed.

[0107] In the embodiment of the present application, the present invention obtains a physically interpretable damage fingerprint matrix through cross-scale physical field coupling modeling, which can accurately characterize the damage feature matrix of the sensor caused by the coupling of multiple physical fields at the micro-nano scale, avoiding the evaluation bias caused by the traditional method ignoring the interaction of multiple physical fields; combining historical accelerated aging data with field operation data, and using dynamic transfer learning driven by neural differential equations to construct a self-evolving digital twin evaluation engine, which can adapt to the performance changes of MEMS thin film acoustic vibration sensors in different working conditions and aging stages in real time, solving the problem of insufficient generalization ability of traditional models; finally, based on the digital twin evaluation engine, with the damage fingerprint matrix and working condition context as input, a dynamic health index and an autonomous maintenance instruction set are generated through multi-agent reinforcement learning decision optimization, realizing dynamic quantitative evaluation of the sensor health status and intelligent maintenance decision-making, significantly improving the accuracy and real-time performance of the MEMS thin film acoustic vibration sensor status evaluation, providing reliable technical support for the precise status monitoring and intelligent maintenance of the sensor, effectively reducing the risk of equipment failure and maintenance costs, and improving the reliability and operation efficiency of the system.

[0108] The present invention constructs a multi-physical field constitutive relationship based on the geometric parameters and material properties of the micro-nano structure of the MEMS thin film acoustic vibration sensor, and designs a cross-scale coupling interface to achieve accurate simulation of the multi-physical field coupling effect at the micro-nano scale of the sensor. The cross-scale physical field coupling model is solved and verified using the actual working environment parameters, thereby obtaining a physically interpretable damage fingerprint matrix. This fully considers the complex damage mechanism of the sensor under the action of different physical fields. Compared with the traditional method that only considers a single physical field or a simple combination, it can more accurately reflect the actual damage state of the sensor, provide a reliable basis for subsequent state evaluation, and greatly improve the accuracy and reliability of state monitoring.

[0109] Based on historical accelerated aging data and field operation data, this paper uses a dynamic transfer learning method driven by neural differential equations to construct a self-evolving digital twin evaluation engine. As field operation data accumulates, the model parameters are dynamically adjusted to gradually reduce the impact of historical accelerated aging data, enabling the evaluation engine to adapt in real time to changes in sensor performance under different operating conditions and aging stages. This dynamic adaptability addresses the lack of generalization capabilities of traditional evaluation models, ensuring that the digital twin evaluation engine can accurately evaluate the status of MEMS thin film acoustic vibration sensors in a variety of complex environments, enhancing its practicality and adaptability, and providing a strong guarantee for the long-term stable operation of MEMS thin film acoustic vibration sensors.

[0110] Based on the damage fingerprint matrix and operating condition context, this invention uses a digital twin assessment engine to perform multi-agent reinforcement learning decision optimization, generating a dynamic health index and an autonomous maintenance instruction set. By setting up a state assessment agent, a maintenance decision agent, and a resource management agent, and defining their state space, action space, and reward function, and utilizing multi-agent policy gradients for decision optimization, it can comprehensively consider multiple factors such as the sensor's current damage state, operating condition information, and maintenance resources to make the optimal maintenance decision. The dynamic health index intuitively reflects the sensor's health status, while the autonomous maintenance instruction set details preventive, corrective, and predictive maintenance operations, achieving closed-loop management of the entire process from condition monitoring to intelligent maintenance. This effectively improves maintenance efficiency, reduces maintenance costs, and enhances the operational stability and reliability of the entire system.

[0111] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0112] Reference Figure 3 , shows a thin film acoustic vibration sensor state assessment device for a micro-electromechanical system provided by an embodiment of the present application, specifically comprising the following modules: A damage fingerprint matrix establishment module 310 is used to perform coupled modeling based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix; A digital twin evaluation engine establishment module 320 is used to perform dynamic transfer learning based on historical accelerated aging data and operating data of the MEMS thin film acoustic vibration sensor to establish a digital twin evaluation engine; A data acquisition module 330 is used to acquire target micro-nanoscale dynamic response data, target field operation data, and target operating condition data corresponding to a target MEMS thin film acoustic vibration sensor; A dynamic health index and autonomous maintenance instruction set module 340 is configured to perform decision optimization on the target micro-nanoscale data, the target field operation data, and the target operating condition data using the damage fingerprint matrix, the digital twin assessment engine, and the operating condition context, to obtain a dynamic health index and an autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor; In one embodiment of the present invention, the damage fingerprint matrix establishment module 310 includes: A multi-physics field constitutive relationship construction submodule is used to construct a multi-physics field constitutive relationship based on the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states; A cross-scale coupling interface design submodule, used to design a cross-scale coupling interface based on the multi-physics constitutive relationship and the interaction mechanism between different physical fields; A verified cross-scale physical field coupling model construction submodule, used to solve and verify the cross-scale model according to the cross-scale coupling interface, and to construct a verified cross-scale physical field coupling model; The damage fingerprint matrix construction submodule is used to calculate the characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states based on the verified cross-scale physical field coupling model, and construct the damage fingerprint matrix with physical interpretability; wherein the characteristic parameters include vibration frequency offset, amplitude attenuation rate and electrical performance change.

[0113] In one embodiment of the present invention, the cross-scale coupling interface design submodule includes: A piezoelectric effect submodule, configured to design a piezoelectric effect coupling interface based on the piezoelectric effect; wherein the piezoelectric effect coupling interface is configured to use the strain generated by the mechanical field as input to the electrical field, and simultaneously use the electric field generated by the electrical field as feedback to the mechanical field; The thermal stress effect submodule is used to design a thermal stress effect coupling interface based on the thermal stress effect; wherein the thermal stress effect coupling interface is used to use the thermal stress caused by the temperature change generated by the thermal field as the input of the mechanical field, and at the same time use the stress generated by the mechanical field as the feedback of the thermal field.

[0114] In one embodiment of the present invention, the digital twin evaluation engine establishment module 320 includes: A neural differential equation model submodule, configured to perform data preprocessing and feature extraction based on the historical accelerated aging data, and construct a neural differential equation model; An aligned operating data submodule, configured to align the operating data with the historical accelerated aging data and perform domain adaptation to obtain aligned operating data; an updated neural differential equation model submodule, configured to perform dynamic transfer learning based on the aligned running data and the neural differential equation model to construct an updated neural differential equation model; The self-evolving digital twin evaluation engine submodule is used to construct a self-evolving digital twin evaluation engine based on the updated neural differential equation model.

[0115] In one embodiment of the present invention, the self-evolving digital twin assessment engine submodule includes: An evaluation result submodule is used to integrate the updated neural differential equation model with the digital twin of the MEMS thin film acoustic vibration sensor, monitor and evaluate the state of the MEMS thin film acoustic vibration sensor in real time, and obtain an evaluation result; A digital twin evaluation engine with self-evolution capability is constructed as a submodule, which is used to feed back the evaluation results to the digital twin of the MEMS thin film acoustic vibration sensor to construct a digital twin evaluation engine with self-evolution capability.

[0116] In one embodiment of the present invention, the dynamic health index and autonomous maintenance instruction set module 340 includes: The feature vector submodule is used to construct a condition assessment agent, a maintenance decision agent, and a resource management agent based on the damage fingerprint matrix and the working condition context into a feature vector; A multi-agent reinforcement learning model submodule is used to obtain an optimized multi-agent reinforcement learning model based on the decision optimization of the state assessment agent, the maintenance decision agent, and the resource management agent based on the digital twin evaluation engine; The decision result generation submodule is used to input the target micro-nanoscale data, the target field operation data and the target working condition data into the optimized multi-agent reinforcement learning model to generate a decision result, a dynamic health index and an autonomous maintenance instruction set.

[0117] In one embodiment of the present invention, the decision result generation submodule includes: A decision result extraction submodule is used to extract the decision result to obtain the current damage level, remaining service life prediction and performance degradation trend; A dynamic health index submodule, configured to generate a dynamic health index based on preset health status quantification rules, the current damage level, the remaining service life prediction, and the performance degradation trend; The autonomous maintenance instruction set submodule is used to obtain an autonomous maintenance instruction set based on the decision result; wherein the autonomous maintenance instruction set includes preventive maintenance, corrective maintenance and predictive maintenance.

[0118] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0119] Reference Figure 4 , shows a computer device for evaluating the status of a thin film acoustic vibration sensor of a micro-electromechanical system of the present application, which can specifically include the following: coupling modeling based on the dynamic response of micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix; dynamic transfer learning is performed based on the historical accelerated aging data and operation data of the MEMS thin film acoustic vibration sensor to establish a digital twin evaluation engine; target micro-nanoscale dynamic response data, target on-site operation data and target working condition data corresponding to the target MEMS thin film acoustic vibration sensor are obtained; decision optimization is performed on the target micro-nanoscale data, the target on-site operation data and the target working condition data through the damage fingerprint matrix, the digital twin evaluation engine and the working condition context to obtain a dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor.

[0120] The computer device 1 is a general-purpose computing device. The components of the computer device 1 may include but are not limited to: one or more processors or processing units 3, memory 8, and a bus 4 connecting different system components (including memory 8 and processing unit 3).

[0121] Bus 4 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Audio Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0122] The computer device 1 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 1, including volatile and non-volatile media, removable and non-removable media.

[0123] The memory 8 may include computer system readable media in the form of volatile memory, such as random access memory 9 and / or cache memory 10. The computer device 1 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 11 may be configured to read and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). Although Figure 4Although not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk"), as well as an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM, or other optical media) can be provided. In these cases, each drive can be connected to bus 4 via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 13. These program modules 13 are configured to perform the functions of various embodiments of the present application.

[0124] A program / utility 12 having a set (at least one) of program modules 13, which may be stored, for example, in a memory, includes, but is not limited to, an operating system, one or more application programs, other program modules 13, and program data, each of which, or some combination thereof, may include an implementation of a network environment. The program modules 13 generally implement the functions and / or methods of the embodiments described herein.

[0125] The computer device 1 may also communicate with one or more external devices 2 (e.g., a keyboard, a pointing device, a display 7, a camera, etc.), one or more devices that enable an operator to interact with the computer device 1, and / or any device that enables the computer device 1 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication may be performed via an I / O interface 6. Furthermore, the computer device 1 may also communicate with one or more networks (e.g., a local area network (LAN)), a wide area network (WAN), and / or a public network (e.g., the Internet) via a network adapter 5. Figure 4 As shown, the network adapter 5 communicates with other modules of the computer device 1 via the bus 4. Figure 4 Not shown, other hardware and / or software modules may be used in conjunction with the computer device 1, including but not limited to: microcode, device drivers, redundant processing units 3, external disk drive arrays, RAID systems, tape drives, and data backup storage systems 11, etc.

[0126] The processing unit 3 executes various functional applications and data processing by running the programs stored in the memory 8, for example, implementing the state evaluation of a thin film acoustic vibration sensor of a micro-electromechanical system provided in an embodiment of the present application.

[0127] That is, when the above-mentioned processing unit 3 executes the above-mentioned program, it realizes: coupling modeling is performed based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix; dynamic transfer learning is performed based on the historical accelerated aging data and operation data of the MEMS thin film acoustic vibration sensor to establish a digital twin evaluation engine; target micro-nanoscale dynamic response data, target field operation data and target working condition data corresponding to the target MEMS thin film acoustic vibration sensor are obtained; decision optimization is performed on the target micro-nanoscale data, the target field operation data and the target working condition data through the damage fingerprint matrix, the digital twin evaluation engine and the working condition context to obtain the dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor.

[0128] In an embodiment of the present application, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a thin film acoustic vibration sensor state evaluation of a micro-electromechanical system as provided in all embodiments of the present application.

[0129] That is, when the program is executed by the processor, it is implemented as follows: coupling modeling is performed based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix; dynamic transfer learning is performed based on the historical accelerated aging data and operation data of the MEMS thin film acoustic vibration sensor to establish a digital twin evaluation engine; target micro-nanoscale dynamic response data, target on-site operation data and target working condition data corresponding to the target MEMS thin film acoustic vibration sensor are obtained; decision optimization is performed on the target micro-nanoscale data, the target on-site operation data and the target working condition data through the damage fingerprint matrix, the digital twin evaluation engine and the working condition context to obtain a dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor.

[0130] Any combination of one or more computer-readable media may be employed. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program for use by or in connection with an instruction execution system, apparatus, or device.

[0131] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0132] The computer program code for performing the operations of the present application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the operator's computer, partially on the operator's computer, as a separate software package, partially on the operator's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the operator's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, through the Internet using an Internet service provider). The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referenced.

[0133] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0134] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0135] The above is a detailed introduction to the thin film acoustic vibration sensor state evaluation method and device for a micro-electromechanical system provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system, characterized in that: include; Based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states, a damage fingerprint matrix is ​​established; Dynamically transfer learning is performed based on historical accelerated aging data and operating data of MEMS thin film acoustic vibration sensors to establish a digital twin evaluation engine; Acquire target micro-nanoscale dynamic response data, target field operation data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor; The target micro-nanoscale data, the target field operation data and the target working condition data are subject to decision optimization through the damage fingerprint matrix, the digital twin evaluation engine and the working condition context, so as to obtain a dynamic health index and an autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor.

2. The method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to claim 1, characterized in that: The step of performing coupled modeling based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish a damage fingerprint matrix includes: Constructing a multi-physics constitutive relationship based on the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states; Designing a cross-scale coupling interface based on the multi-physics constitutive relations and the interaction mechanisms between different physical fields; Solve and verify the cross-scale model according to the cross-scale coupling interface to build a verified cross-scale physical field coupling model; The characteristic parameters of the MEMS thin film acoustic vibration sensor under different damage states are calculated based on the verified cross-scale physical field coupling model, and the damage fingerprint matrix with physical interpretability is constructed; wherein the characteristic parameters include vibration frequency offset, amplitude attenuation rate and electrical performance change.

3. The method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to claim 2, characterized in that: The multi-physics constitutive relationship includes piezoelectric effect and thermal stress effect, the cross-scale coupling interface includes a piezoelectric effect coupling interface and a thermal stress effect coupling interface, and the step of designing the cross-scale coupling interface based on the multi-physics constitutive relationship and the interaction mechanism between different physical fields includes: A piezoelectric effect coupling interface is designed based on the piezoelectric effect; wherein the piezoelectric effect coupling interface is used to use the strain generated by the mechanical field as the input of the electric field, and simultaneously use the electric field generated by the electric field as the feedback of the mechanical field; A thermal stress effect coupling interface is designed based on the thermal stress effect; wherein the thermal stress effect coupling interface is used to use the thermal stress caused by the temperature change generated by the thermal field as the input of the mechanical field, and at the same time use the stress generated by the mechanical field as the feedback of the thermal field.

4. The method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to claim 1, wherein: The step of performing dynamic transfer learning based on historical accelerated aging data and operating data of the MEMS thin film acoustic vibration sensor to establish a digital twin evaluation engine includes: Performing data preprocessing and feature extraction based on the historical accelerated aging data to construct a neural differential equation model; Aligning the operating data with the historical accelerated aging data and adapting them to the domain to obtain aligned operating data; Performing dynamic transfer learning based on the aligned operating data and the neural differential equation model to construct an updated neural differential equation model; A self-evolving digital twin evaluation engine is constructed based on the updated neural differential equation model.

5. The method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to claim 4, characterized in that: The step of constructing a self-evolving digital twin evaluation engine based on the updated neural differential equation model includes: Establishing a digital twin of a MEMS thin film acoustic vibration sensor, integrating the updated neural differential equation model with the digital twin of the MEMS thin film acoustic vibration sensor, and monitoring and evaluating the state of the MEMS thin film acoustic vibration sensor in real time to obtain an evaluation result; The evaluation results are fed back to the digital twin of the MEMS thin film acoustic vibration sensor to build a digital twin evaluation engine with self-evolution capability.

6. The method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to claim 1, characterized in that: The step of performing decision optimization on the target micro-nanoscale data, the target field operation data, and the target operating condition data through the damage fingerprint matrix, the digital twin assessment engine, and the operating condition context to obtain a dynamic health index and an autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor includes: According to the damage fingerprint matrix and the working condition context, a feature vector is spliced ​​to construct a state assessment agent, a maintenance decision agent and a resource management agent; Obtaining an optimized multi-agent reinforcement learning model based on the decision optimization of the state assessment agent, the maintenance decision agent, and the resource management agent based on the digital twin evaluation engine; The target micro-nanoscale data, the target on-site operation data, and the target working condition data are input into the optimized multi-agent reinforcement learning model to generate a decision result, and a dynamic health index and an autonomous maintenance instruction set are generated.

7. The method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to claim 6, characterized in that: The step of inputting the target micro-nanoscale data, the target field operation data, and the target operating condition data into the optimized multi-agent reinforcement learning model to generate a dynamic health index and an autonomous maintenance instruction set includes: Extracting the decision results to obtain the current damage level, remaining service life prediction and performance degradation trend; generating a dynamic health index according to a preset health status quantification rule, the current damage level, the remaining service life prediction, and the performance degradation trend; An autonomous maintenance instruction set is obtained according to the decision result; wherein the autonomous maintenance instruction set includes preventive maintenance, corrective maintenance and predictive maintenance.

8. A thin film acoustic vibration sensor state evaluation device for a micro-electromechanical system, characterized in that: include: The damage fingerprint matrix establishment module is used to conduct coupled modeling based on the dynamic response of the micro-nanoscale data of the MEMS thin film acoustic vibration sensor under different damage states to establish the damage fingerprint matrix; A digital twin evaluation engine establishment module is used to perform dynamic transfer learning based on historical accelerated aging data and operating data of MEMS thin film acoustic vibration sensors to establish a digital twin evaluation engine; A data acquisition module is used to acquire target micro-nanoscale dynamic response data, target field operation data, and target operating condition data corresponding to the target MEMS thin film acoustic vibration sensor; The dynamic health index and autonomous maintenance instruction set module is used to optimize the decision-making of the target micro-nanoscale data, the target field operation data and the target operating condition data through the damage fingerprint matrix, the digital twin evaluation engine and the operating condition context, and obtain the dynamic health index and autonomous maintenance instruction set corresponding to the target MEMS thin film acoustic vibration sensor.

9. A computer electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the method implements the steps of the method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for evaluating the state of a thin film acoustic vibration sensor of a micro-electromechanical system according to any one of claims 1 to 7 are implemented.

Images