A MEMS vibration sensor resonant frequency self-detection method, system and device

Abstract

The present invention discloses a self-check method, system and device for the resonant frequency of a MEMS vibration sensor, the method comprising the following steps: using a time domain continuous signal excitation as a response input signal, and collecting a dynamic response signal of the MEMS vibration sensor; establishing a dynamic response mathematical model of the MEMS vibration sensor; using the collected dynamic response signal as identification data, using a BP neural network optimized by an improved particle swarm algorithm to perform high-precision identification of the resonant frequency parameters in the dynamic response mathematical model; and calculating the mechanical sensitivity of the MEMS vibration sensor based on the identification result of the resonant frequency parameters. The present invention relates to the technical field of dynamic model parameter identification of MEMS vibration sensors, and can effectively improve the identification accuracy of the resonant frequency of MEMS vibration sensors.

Description

Technical Field

[0001] The invention relates to the technical field of dynamic model parameter identification of a MEMS vibration sensor, and in particular to a self-checking method, system and equipment for a resonant frequency of a MEMS vibration sensor. Background Art

[0002] MEMS vibration sensors have a large range, small size, high bandwidth, flexible interface matching and high robustness, and are widely used in attitude detection, inertial navigation, vibration monitoring and other fields. Since the sensitive structure of MEMS vibration sensors has quality factor and gain requirements, such as the amplitude gain coefficient of butterfly-wing sensors is not less than 10dB, the resonant frequency detection of sensitive structures can assist in the detection and calibration of the mechanical sensitivity of MEMS vibration sensors. Therefore, improving the accuracy of MEMS vibration sensor resonant frequency identification is of great significance to the practical application of MEMS vibration sensors.

[0003] The classic vibration sensor dynamic model parameter identification strategy can be divided into two categories: time domain and frequency domain. The time domain method strictly follows ISO 16063-12:20011 "Absolute method shock excitation calibration standard", accurately captures the real-time response data of the vibration sensor system under standard pulse shock, and cleverly uses advanced methods such as least squares method to achieve accurate estimation of model parameters. In the frequency domain, based on ISO 16063-11:1999 "Calibration method for vibration and shock sensors", with the deep integration of steady-state sinusoidal vibration testing and laser interferometry measurement technology, the detailed frequency response function of the vibration sensor is derived using fine analysis methods such as sine approximation method, fringe counting method and minimum point method.

[0004] In the process of in-depth exploration of this technical field, the following problems were found. First, the identification process of the dynamic model of the vibration sensor is inevitably affected by the complex noise environment of the system, which significantly increases the difficulty of identification. Secondly, traditional model construction mostly relies on the description of the inherent parameters of the system. However, since some inherent parameters themselves have low sensitivity characteristics, their identification accuracy is easily affected by the negative impact of system noise, which restricts the further development space of traditional methods in improving the identification accuracy of the dynamic model parameters of vibration sensors. In view of this, exploring and implementing more accurate and efficient identification strategies to effectively respond to the above challenges has become a core research topic and urgent need in the current field. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method, system and device for self-detection of the resonant frequency of a MEMS vibration sensor, which can effectively improve the recognition accuracy of the resonant frequency of the MEMS vibration sensor.

[0006] To achieve the above object, the present invention provides a MEMS vibration sensor resonant frequency self-test method, comprising the following steps:

[0007] Step 1, using a time domain continuous signal excitation as a response input signal, and collecting a dynamic response signal of a MEMS vibration sensor;

[0008] Step 2, establishing a dynamic response mathematical model of the MEMS vibration sensor;

[0009] Step 3, using the collected dynamic response signal as identification data, using the BP neural network optimized by the improved particle swarm algorithm to perform high-precision identification of the resonant frequency parameters in the dynamic response mathematical model, specifically:

[0010] Firstly, the threshold and weight of BP neural network are randomly generated as the initial particle position of the improved particle swarm optimization algorithm, and the algorithm related parameters are set;

[0011] Then, the cost function of the BP neural network is set as the fitness function of the improved particle swarm algorithm, and the position and velocity of the particles are updated through iteration, and iteration is performed with the goal of minimizing the fitness until the accuracy requirement is met or the maximum number of iterations is reached;

[0012] Then, the global optimal solution of the improved particle swarm optimization algorithm is used as output to construct a discrete transfer function of the dynamic response mathematical model, and further perform discrete-to-continuous conversion to calculate the resonant frequency parameters;

[0013] Step 4: Calculate the mechanical sensitivity of the MEMS vibration sensor based on the identification result of the resonant frequency parameter.

[0014] In one embodiment, in step 2, the dynamic response mathematical model is specifically:

[0015] ;

[0016] in, H ( s ) is the dynamic response mathematical model function, s is the Laplace variable, b , a 1. a 0 is the parameter to be identified;

[0017] Specifically:

[0018] ;

[0019] in, is the effective moment of inertia of the sensor, J is the rotational inertia of the sensitive structure, Qis the sensor quality factor, is the resonant frequency, , J , Q , This is the resonant frequency parameter in step 3.

[0020] In one embodiment, the cost function of the improved particle swarm optimization algorithm is H The mean square error criterion is adopted, that is:

[0021] ;

[0022] in, N is the number of input signal acquisition points, E (·) is the mathematical expectation.

[0023] In one embodiment, the improved particle swarm optimization algorithm adopts an adaptive inertia weight, namely:

[0024] ;

[0025] in, w is the inertia coefficient, To preset the minimum inertia coefficient, is the preset maximum inertia coefficient, is the mean fitness of all particles in the current iteration, is the lowest fitness among all particles in the current iteration. F is the fitness of the current particle.

[0026] In one embodiment, the learning factor adopted by the improved particle swarm optimization algorithm changes nonlinearly, that is:

[0027] ;

[0028] in, c 1 is the individual learning factor, c 2 is the global learning factor, To preset the maximum learning factor, To preset the minimum learning factor, k is the number of iterations, K The preset maximum number of iterations.

[0029] In one embodiment, in step 5, the calculating of the mechanical sensitivity of the MEMS vibration sensor is specifically:

[0030] ;

[0031] in, S is the mechanical sensitivity of the MEMS vibration sensor.

[0032] To achieve the above object, the present invention further provides a MEMS vibration sensor resonant frequency self-test system, using the above method, the resonant frequency self-test system comprises:

[0033] A response signal generating unit, used for adopting a time domain continuous signal excitation as a response input signal and collecting a dynamic response signal of the MEMS vibration sensor;

[0034] A model building unit, used to establish a dynamic response mathematical model of the MEMS vibration sensor;

[0035] A parameter identification unit, using the collected dynamic response signal as identification data, and adopting a BP neural network optimized by an improved particle swarm algorithm to perform high-precision identification of the resonant frequency parameters in the dynamic response mathematical model;

[0036] The sensitivity calculation unit calculates the mechanical sensitivity of the MEMS vibration sensor based on the identification result of the resonance frequency parameter.

[0037] To achieve the above object, the present invention further provides a terminal device, wherein the terminal device is provided with:

[0038] Memory, used to store programs;

[0039] The processor is used to execute the program stored in the memory. When the program is executed, the processor is used to execute the above method.

[0040] Compared with the prior art, the present invention has the following beneficial technical effects:

[0041] The present invention innovatively proposes a high-precision self-test technology for the resonant frequency of MEMS vibration sensors. The core of this technology is to use the BP neural network model finely tuned by the improved particle swarm algorithm to achieve accurate modeling and parameter identification of the dynamic characteristics of the MEMS vibration sensor. In particular, this technology focuses on the resonant frequency parameters in the standard dynamic model of the sensor. Through the integration of deep learning and optimization technology, it significantly improves the identification accuracy of this key parameter and effectively shields the interference of environmental noise on the estimation of system model parameters. In addition, this method also indirectly promotes a leap in the accuracy of dynamic model parameter identification of related equipment such as digital vibration sensors, bringing more reliable data support and technical guarantees to the field of vibration measurement and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0043] Figure 1 This is a flow chart of the self-test method for the resonant frequency of a MEMS vibration sensor in Example 1 of the present invention;

[0044] Figure 2 This is a schematic diagram of the input signal of the MEMS vibration sensor in Example 1 of the present invention;

[0045] Figure 3 This is a schematic diagram of the response data of the MEMS vibration sensor in Example 1 of the present invention;

[0046] Figure 4 This is a schematic diagram of the input and output of the BP neural network in Example 1 of the present invention;

[0047] Figure 5 This is a flow chart of improving the particle swarm algorithm to optimize the BP neural network in Example 1 of the present invention;

[0048] Figure 6 Schematic diagram of the dynamic model identification result of the MEMS vibration sensor in Example 1 of the present invention, wherein: Figure 6 (a) is a schematic diagram of the training fitting effect. Figure 6 (b) is a schematic diagram of the identification results. Figure 6 (c) is a schematic diagram of the training set error. Figure 6 (d) is a schematic diagram of identification error;

[0049] Figure 7 Schematic diagram of the cumulative error curve of the dynamic model identification result in Example 1 of the present invention, wherein: Figure 7 (a) is a schematic diagram of the weight and threshold update curve. Figure 7 (b) is a schematic diagram of particle velocity. Figure 7 (c) is a schematic diagram of the cumulative error iteration curve;

[0050] Figure 8 It is a structural block diagram of the resonant frequency self-test system of the MEMS vibration sensor in Example 2 of the present invention;

[0051] Fig. 9 This is a structural block diagram of the terminal device in Example 3 of the present invention.

[0052] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0055] Example 1

[0056] like Figure 1 The figure shows a MEMS vibration sensor resonant frequency self-test method disclosed in this embodiment, which mainly includes the following steps:

[0057] Step 1, using a time domain continuous signal excitation as a response input signal, and collecting a dynamic response signal of a MEMS vibration sensor;

[0058] Step 2, establishing a dynamic response mathematical model of the MEMS vibration sensor;

[0059] Step 3, using the collected dynamic response signal as identification data, and using the BP neural network optimized by the improved particle swarm algorithm to perform high-precision identification of the resonant frequency parameters in the dynamic response mathematical model;

[0060] Step 4: Calculate the mechanical sensitivity of the MEMS vibration sensor based on the identification result of the resonant frequency parameter.

[0061] In the specific implementation process of step 1, a time-domain continuous voltage excitation is used as the response input signal of the system, and a standard excitation signal is generated internally or externally. Figure 2As shown, the standard pulse signal generated by the built-in MCU of the butterfly-wing MEMS vibration sensor is used in this example. In addition, the principle of ensuring data quality and adequacy must be strictly followed, with the goal of accurately capturing the vibration characteristics of the MEMS vibration sensor. Specifically, a minimum standard for the number of discrete signal sampling points is set, requiring that the standard should not only meet the stringent requirements of subsequent complex data analysis and advanced algorithm processing, but also ensure that the collected time-domain dynamic response signal can fully and completely reflect the vibration behavior of the sensor. In this embodiment, 2000 sampling points are used as sample data. This value far exceeds the minimum threshold required, which lays a solid data foundation for subsequent model construction, parameter identification and performance evaluation, and effectively guarantees the accuracy and reliability of the entire analysis process.

[0062] In the specific implementation process of step 2, the dynamic response mathematical model of the butterfly-wing MEMS vibration sensor is specifically constructed as follows:

[0063] (1)

[0064] in, H ( s ) is the dynamic response mathematical model function, s is the Laplace variable, b , a 1. a 0 is the parameter to be identified;

[0065] The parameters to be identified in formula (1) are specifically expressed as:

[0066] (2)

[0067] in, is the effective moment of inertia of the sensor, J is the rotational inertia of the sensitive structure, Q is the sensor quality factor, is the resonant frequency, , J , Q , This is the resonant frequency parameter in step 3.

[0068] It is worth noting that, in the specific application process, the dynamic response mathematical model of the MEMS vibration sensor is not limited to the form of formula (1) and formula (2), and can also adopt other forms including the resonant frequency parameters to be identified.

[0069] In the specific implementation process of step 3, in order to enhance the efficiency and effect of the improved particle swarm algorithm, the mean square error is used as the core indicator for evaluating the consistency between the model prediction performance and the actual data. H ,Right now:

[0070] (3)

[0071] in, N is the number of input signal acquisition points, E (·) is the mathematical expectation.

[0072] Compared with the error mean and standard deviation, the mean square error is preferred in the pursuit of efficient evaluation scenarios because of its directness in calculation and high sensitivity to the error size. The mean square error amplifies the impact of large errors through the sum of squares, allowing the algorithm to more accurately locate and optimize these large deviations during the iteration process, thereby accelerating the convergence process to the global optimal solution.

[0073] In addition, in the improved particle swarm algorithm, a fixed inertia weight is traditionally used w Maintain the inertia of particle motion. However, as the number of iterations increases, the complexity of the problem changes, and the fixed w It is difficult to adapt, which may lead to problems such as slow convergence and local optimality. In order to overcome the above disadvantages, a variable inertia weight strategy is proposed, namely:

[0074] (4)

[0075] in, To preset the minimum inertia coefficient, is the preset maximum inertia coefficient, is the mean fitness of all particles in the current iteration, is the lowest fitness among all particles in the current iteration. F Is the fitness of the current particle. Dynamically adjusted according to the iterative process w Value, initial w Greatly promote global exploration, later w Smaller acceleration of local convergence, balanced search capabilities, and improved algorithm adaptability and solution efficiency.

[0076] Further preferably, in the particle swarm optimization algorithm, the individual learning factor and the global learning factor are dynamically adjusted, and the individual learning factor is increased in the initial stage. c 1. Increase the global learning factor in the later stage c 2. The algorithm can be improved. However, the linear adjustment strategy is insufficient and does not fully meet the algorithm requirements. A nonlinear adjustment strategy is proposed, which uses the sigmoid function to dynamically set the learning factor value, namely:

[0077] (5)

[0078] in, To preset the maximum learning factor, To preset the minimum learning factor, k is the number of iterations,K The maximum number of iterations is preset. Slow changes are made in the early and late stages, and rapid adjustments are made in the middle stage to balance global and local searches and improve the algorithm's adaptability and optimization performance.

[0079] refer to Figure 4 ,BP neural network is a multi-layer feedforward network, which is trained by error back propagation algorithm. Figure 4 middle, u ( k -2) is the input second-order differential signal, y ( k -1) is the output first-order differential signal, y ( k -2) is the output second-order differential signal, To predict the output signal. The structure of the BP neural network includes an input layer, a hidden layer (multiple layers are possible) and an output layer. The key points are: forward propagation of samples, back propagation of errors to adjust weights, and iterative training to approximate the expected output. This embodiment uses a single hidden layer neural network, and the number of nodes is set to 8.

[0080] refer to Figure 5 , the BP neural network optimized by improved particle swarm algorithm is used to identify the resonant frequency parameters in the dynamic response mathematical model with high precision. Specifically:

[0081] Firstly, the threshold and weight of BP neural network are randomly generated as the initial particle position of the improved particle swarm optimization algorithm, and the algorithm related parameters are set;

[0082] Then, the cost function of the BP neural network is set as the fitness function of the improved particle swarm algorithm, and the position and velocity of the particles are updated through iteration, and iteration is performed with the goal of minimizing the fitness until the accuracy requirement is met or the maximum number of iterations is reached;

[0083] Then, the global optimal solution of the improved particle swarm optimization algorithm is taken as the output to construct the discrete transfer function of the dynamic response mathematical model, and further discrete-to-continuous conversion is performed to calculate the resonant frequency parameters.

[0084] refer to Figure 6 , Figure 7The model identification results presented here show that the optimized BP neural network exhibits good identification performance. Specifically, its identification accuracy is significantly improved and the cumulative error is greatly reduced, which shows that the network has higher accuracy and stability when processing complex data patterns. It is particularly worth mentioning that in the identification of resonant frequency, the network shows extremely small errors and achieves very high identification accuracy, which to a certain extent effectively overcomes the common error effects of traditional methods when dealing with such problems. In summary, the optimized BP neural network provides solid data support for the accurate modeling and performance evaluation of vibration sensor models.

[0085] In the specific implementation process, the mechanical sensitivity of the MEMS vibration sensor is calculated as follows:

[0086] ;

[0087] in, S is the mechanical sensitivity of the MEMS vibration sensor.

[0088] It is worth noting that although this embodiment Figure 1 , Figure 5 The steps in the process are shown in sequence as indicated by the arrows, but the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of the steps, and the steps can be executed in other orders. Figure 1 , Figure 5 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0089] Example 2

[0090] Based on the MEMS vibration sensor resonant frequency self-test method in Example 1, this embodiment discloses a MEMS vibration sensor resonant frequency self-test system, referring to Figure 8 The MEMS vibration sensor resonant frequency self-test system includes a response signal generating unit, a model building unit, a parameter identification unit and a sensitivity calculation unit. Specifically:

[0091] The response signal generating unit is used to use a time domain continuous signal excitation as a response input signal and collect a dynamic response signal of the MEMS vibration sensor;

[0092] The model building unit is used to establish a dynamic response mathematical model of the MEMS vibration sensor;

[0093] The parameter identification unit uses the collected dynamic response signal as the identification data, and uses the BP neural network optimized by the improved particle swarm algorithm to perform high-precision identification of the resonant frequency parameters in the dynamic response mathematical model;

[0094] The sensitivity calculation unit calculates the mechanical sensitivity of the MEMS vibration sensor based on the identification result of the resonance frequency parameter.

[0095] In this embodiment, the specific working process and working principle of the response signal generating unit, the model building unit, the parameter identification unit and the sensitivity calculation unit are the same as those in the method of Embodiment 1, so they are not described in detail in this embodiment. Each unit module can be implemented in whole or in part by software, hardware and a combination thereof, and each unit module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above unit modules.

[0096] Example 3

[0097] like Fig. 9 The terminal device disclosed in this embodiment includes a transmitter, a receiver, a memory, and a processor. The transmitter is used to send instructions and data, the receiver is used to receive instructions and data, the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions stored in the memory to implement the method in the above-mentioned embodiment 1.

[0098] It should be noted that the above memory can be independent or integrated with the processor. When the memory is independently provided, the terminal device further includes a bus for connecting the memory and the processor.

[0099] Example 4

[0100] This embodiment discloses a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the method in the above-mentioned embodiment 1 is implemented.

[0101] The above description is only a preferred embodiment of the present invention, and does not limit the scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the protection scope of the present invention.

Claims

1. A self-test method for resonant frequency of a MEMS vibration sensor, characterized in that: The following steps are involved: Step 1, using a time domain continuous signal excitation as a response input signal, and collecting a dynamic response signal of a MEMS vibration sensor; Step 2: Establish a dynamic response mathematical model of the MEMS vibration sensor. The dynamic response mathematical model is specifically: ； in, H ( s ) is the dynamic response mathematical model function, s is the Laplace variable, b , a 1. a 0 is the parameter to be identified; Specifically: ； in, is the effective moment of inertia of the sensor, J is the rotational inertia of the sensitive structure, Q is the sensor quality factor, is the resonant frequency, , J , Q , That is the resonant frequency parameter in step 3; Step 3, using the collected dynamic response signal as identification data, using the BP neural network optimized by the improved particle swarm algorithm to perform high-precision identification of the resonant frequency parameters in the dynamic response mathematical model, specifically: Firstly, the threshold and weight of BP neural network are randomly generated as the initial particle position of the improved particle swarm optimization algorithm, and the algorithm related parameters are set; Then, the cost function of the BP neural network is set as the fitness function of the improved particle swarm algorithm, and the position and velocity of the particles are updated through iteration, and iteration is performed with the goal of minimizing the fitness until the accuracy requirement is met or the maximum number of iterations is reached; Then, the global optimal solution of the improved particle swarm optimization algorithm is used as output to construct a discrete transfer function of the dynamic response mathematical model, and further perform discrete-to-continuous conversion to calculate the resonant frequency parameters; Step 4: Calculate the mechanical sensitivity of the MEMS vibration sensor based on the identification result of the resonant frequency parameter.

2. The MEMS vibration sensor resonant frequency self-test method according to claim 1, characterized in that: The cost function of the improved particle swarm optimization algorithm H The mean square error criterion is adopted, that is: ； in, N is the number of input signal acquisition points, E (·) is the mathematical expectation.

3. The MEMS vibration sensor resonant frequency self-test method according to claim 1, characterized in that: The improved particle swarm optimization algorithm adopts adaptive inertia weight, namely: ； in, is the inertia coefficient, To preset the minimum inertia coefficient, is the preset maximum inertia coefficient, is the mean fitness of all particles in the current iteration, is the lowest fitness among all particles in the current iteration. F is the fitness of the current particle.

4. The MEMS vibration sensor resonant frequency self-test method according to claim 1, characterized in that: The learning factor adopted by the improved particle swarm optimization algorithm changes nonlinearly, namely: ； in, c 1 is the individual learning factor, c 2 is the global learning factor, To preset the maximum learning factor, To preset the minimum learning factor, k is the number of iterations, K The preset maximum number of iterations.

5. The MEMS vibration sensor resonant frequency self-test method according to claim 1, characterized in that: In step 5, the calculation of the mechanical sensitivity of the MEMS vibration sensor is specifically as follows: ； in, S is the mechanical sensitivity of the MEMS vibration sensor, is the effective moment of inertia of the sensor, J is the rotational inertia of the sensitive structure, is the resonant frequency.

6. A MEMS vibration sensor resonant frequency self-test system, characterized in that: According to the method described in any one of claims 1 to 5, the resonant frequency self-detection system comprises: A response signal generating unit, used for adopting a time domain continuous signal excitation as a response input signal and collecting a dynamic response signal of the MEMS vibration sensor; A model building unit, used to establish a dynamic response mathematical model of the MEMS vibration sensor; A parameter identification unit, using the collected dynamic response signal as identification data, and adopting a BP neural network optimized by an improved particle swarm algorithm to perform high-precision identification of the resonant frequency parameters in the dynamic response mathematical model; The sensitivity calculation unit calculates the mechanical sensitivity of the MEMS vibration sensor based on the identification result of the resonance frequency parameter.

7. A terminal device, characterized in that: The terminal device is provided with: Memory, used to store programs; A processor, configured to execute the program stored in the memory; when the program is executed, the processor is configured to execute the method according to any one of claims 1 to 5.

Images