Spring operating mechanism fault detection and optimization method and related device

Abstract

The invention discloses a spring operating mechanism fault detection and optimization method and a related device, and the method comprises the steps: carrying out the real-time collection of key data of a spring operating mechanism through a stress sensor, a speed sensor and a vibration sensor; based on the collected key data, constructing a multi-dimensional data framework to obtain a multi-dimensional data set; fault feature extraction and data analysis are carried out on the obtained multi-dimensional data set, an adaptive fault model is constructed based on the extracted fault features, and then a real-time fault detection result is obtained through multi-modal data fusion; and optimizing a stress sensor, a speed sensor and a vibration sensor by utilizing automatic calibration and optimization of reinforcement learning according to the real-time fault detection result. Compared with a traditional fault detection method, high-precision real-time monitoring and fault prediction of the spring operating mechanism are achieved through the multi-modal data fusion and deep learning technology, and comprehensive and reliable technical support is provided for safe operation and maintenance optimization of equipment.

Description

Technical Field

[0001] The invention belongs to the technical field of power equipment fault detection, and in particular relates to a spring operating mechanism fault detection and optimization method and a related device. Background Art

[0002] In the field of fault detection of CT20 spring operating mechanism, it is of great significance to improve the accuracy and real-time performance of detection. Traditional fault detection methods usually rely on regular maintenance and manual inspection. However, these methods have many limitations and defects. For example, it is difficult for traditional methods to monitor the operating status of CT20 spring operating mechanism in real time, and it is easy to have the problem of false alarm or missed faults; secondly, when detecting the operating status of spring operating mechanism, only the data of individual components are collected through a single type of sensor, which makes the detection results not comprehensive enough and unable to accurately determine the type and level of faults. To this end, it is particularly important to improve the fault detection accuracy and adaptive detection ability of spring operating mechanism, realize real-time detection and fault prediction of CT20 spring operating mechanism, more comprehensively capture the operating status of equipment, and reduce false alarm and missed faults. Summary of the invention

[0003] The purpose of the present invention is to provide a spring operating mechanism fault detection and optimization method and related devices to solve the technical defects that the current reliance on a single data source makes the detection results incomplete, and cannot achieve real-time monitoring and fault prediction, resulting in false alarms and missed alarms.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, a spring operating mechanism fault detection and optimization method is provided, comprising: Use stress sensors, velocity sensors and vibration sensors to collect key data of the spring operating mechanism in real time; Based on the collected key data, a multidimensional data framework is constructed to obtain a multidimensional data set; Fault feature extraction and data analysis are performed from the obtained multidimensional data set, and an adaptive fault model is constructed based on the extracted fault features, and then a real-time fault detection result is obtained through multimodal data fusion; The stress sensor, velocity sensor and vibration sensor are optimized through the real-time fault detection results and automatic calibration and optimization using reinforcement learning.

[0005] Furthermore, before the key data of the spring operating mechanism is collected by using the stress sensor, the method further includes: Constructing a geometric model of a spring operating mechanism, defining material properties of various components in the geometric model of the spring operating mechanism, and simulating stress distribution and deformation behavior of the spring operating mechanism under working load by using finite element analysis; According to the actual working state of the spring operating mechanism, the load conditions and boundary conditions are set, and the geometric model of the spring operating mechanism is meshed to generate a preliminary stress distribution cloud map; Using an adaptive stress perception algorithm, the stress distribution data on the stress distribution cloud map is analyzed to automatically identify stress concentration areas and obtain the gradient value of each point in the stress distribution and the area with significant stress changes; The identified stress concentration points are clustered so that adjacent high stress gradient points are combined into a stress concentration area; The ASSA algorithm is used to refine the boundary of the stress concentration area and determine the intensive stress distribution area. The genetic optimization algorithm is used to determine the layout position of the stress sensor on the spring operating mechanism.

[0006] Furthermore, before the key data of the spring operating mechanism is collected by the speed sensor, the method further includes: Using the arranged strain sensors, the initial motion data of the spring operating mechanism under different operating states are obtained, and based on the initial motion data, a motion model of the spring in the spring operating mechanism is established to simulate the dynamic behavior of the spring under static, moving and different operating conditions; Using multimodal motion analysis to capture the motion trajectory of the spring in the spring operating mechanism in real time, obtain visual data, and extract key motion features from the visual data; Utilizing the established spring motion model, extracting the motion parameters of the spring, and using a Kalman filter to fuse the visual data and the motion parameters; The arrangement position of the speed sensor on the spring operating mechanism is determined by a genetic optimization algorithm.

[0007] Furthermore, before the key data of the spring operating mechanism is collected by the vibration sensor, the method further includes: Use existing vibration sensors or temporarily deployed sensors to collect vibration data of the spring operating mechanism under different operating conditions; Combining vibration data with stress data and velocity data, and using multimodal fusion technology, the dynamic behavior of the spring operating mechanism under different operating conditions is analyzed; Using multimodal motion analysis algorithms, the vibration modes of the spring in different operating modes are identified; The vibration patterns under different operating modes are compared to determine the operating conditions with the most severe vibration, and the layout position of the vibration sensor on the spring operating mechanism is determined by a genetic optimization algorithm.

[0008] Furthermore, based on the collected key data, a multidimensional data framework is constructed to obtain a multidimensional data set, which specifically includes: After the stress sensor, velocity sensor and vibration sensor are arranged, a multi-frequency adaptive sampling algorithm is used to dynamically adjust the sampling frequency of key data according to the arrangement positions of the stress sensor, velocity sensor and vibration sensor, so that stress, velocity and vibration multi-dimensional data can be collected simultaneously; A multidimensional data framework is constructed on the stress, velocity and vibration multidimensional data collected simultaneously, and the collected stress, velocity and vibration multidimensional data are standardized and integrated by using multimodal data fusion technology; Integrate the standardized multi-dimensional data into a three-dimensional data matrix, wherein the first dimension is time, the second dimension is sensor position, and the third dimension is data type; Based on the three-dimensional data matrix, a multidimensional data table is generated to obtain a multidimensional data set.

[0009] Furthermore, fault feature extraction and data analysis are performed from the obtained multidimensional data set, and an adaptive fault model is constructed based on the extracted fault features. Then, a real-time fault detection result is obtained through multimodal data fusion, specifically including: Extract stress data, velocity data and vibration data fault features from the multidimensional data set, and use the Pearson correlation coefficient to measure the linear correlation between different fault feature data; The correlation calculation structures corresponding to all fault features are combined into a correlation matrix to identify fault features with mutual correlation and fault features containing redundant information; The random forest algorithm is used to evaluate the importance of each fault feature in fault prediction, and the most important fault features are selected based on the fault feature importance evaluation structure to optimize the fault prediction effect; Based on the extracted fault features, an adaptive fault model is established, and the probability of fault occurrence is calculated in real time using the adaptive fault model; Through multimodal data fusion technology, stress data, velocity data and vibration data are comprehensively analyzed to obtain fault detection results and warning information.

[0010] Furthermore, the stress sensor, velocity sensor and vibration sensor are optimized by using the real-time fault detection result and automatic calibration and optimization of reinforcement learning, specifically including: Based on real-time fault detection results and warning information, automatic calibration and optimization of stress sensors, velocity sensors, and vibration sensors are automatically optimized using reinforcement learning.

[0011] In a second aspect, a spring operating mechanism fault detection and optimization system is provided, comprising: The acquisition module is used to collect key data of the spring operating mechanism in real time; A framework building module, used to build a multidimensional data framework based on the collected key data to obtain a multidimensional data set; A fault feature extraction module, used for extracting fault features from the obtained multi-dimensional data set; Data analysis module, used for data analysis of multidimensional data sets; A model building module for building an adaptive fault model; Reinforcement learning module for optimizing strain, velocity, and vibration sensors.

[0012] In a third aspect, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable in the processor, wherein the processor implements the steps of the spring operating mechanism fault detection and optimization method as described above when executing the computer program.

[0013] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the spring operating mechanism fault detection and optimization method as described above is implemented.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. When performing fault detection on the spring operating mechanism, stress sensors, velocity sensors and vibration sensors that are closely related to the operation of the spring operating mechanism are used to collect data on the operation of the spring operating mechanism in real time, which can quickly extract fault data, thereby realizing rapid detection and warning of spring operating mechanism faults; secondly, in the fault detection process, by combining three different types of collected data, the comprehensiveness and accuracy of the detection results are further improved, avoiding the problems of false alarms and missed alarms; finally, combined with the adaptive fault model, multimodal motion analysis algorithm and automatic calibration and optimization based on reinforcement learning, key data can be acquired and analyzed in real time during the operation of the equipment, greatly improving the timeliness of fault detection, and at the same time optimizing the sensor to ensure reliability during the data acquisition process.

[0015] 2. By constructing an accurate geometric model of the spring operating mechanism and defining the material properties of each component in detail, a solid foundation is provided for the subsequent finite element analysis, ensuring the accuracy and reliability of the simulation results; at the same time, the use of finite element analysis to simulate the stress distribution and deformation behavior of the spring operating mechanism under working load can intuitively display the stress state of the mechanism under different working conditions, which helps to discover potential stress concentration areas and deformation trends in advance, and provide an important reference for preventing failures. Setting load conditions and boundary conditions according to the actual working state of the spring operating mechanism makes the simulation results closer to the actual situation, improving the pertinence and practicality of the analysis; and, through the adaptive stress perception algorithm, the data on the stress distribution cloud map is deeply analyzed, and the stress concentration area and the area with significant stress changes are automatically identified, which not only improves the accuracy and efficiency of stress identification, but also reduces human intervention and errors, providing a more accurate target area for subsequent sensor layout.

[0016] 3. The spring motion model established based on the initial motion data can simulate the dynamic behavior of the spring from static, motion and different operating conditions, which helps to deeply understand the motion characteristics of the spring and provides theoretical support for subsequent data collection and analysis; the use of multimodal motion analysis technology can capture the motion trajectory of the spring in the spring operating mechanism in real time, and obtain rich visual data, extract key motion features, speed, acceleration, etc. from the visual data, and provide multi-dimensional information for data analysis. The use of Kalman filter to fuse visual data and motion parameters extracted from the spring motion model can effectively reduce data noise and errors and improve data accuracy and reliability.

[0017] 4. Use existing vibration sensors or temporarily arranged sensors to collect vibration data of the spring operating mechanism under different operating conditions, which provides a rich data source for subsequent data analysis, making the analysis results more comprehensive and accurate. Combining vibration data with stress data and velocity data and analyzing them through multimodal fusion technology can comprehensively consider the interactions and influences between different physical quantities, thereby more accurately revealing the dynamic behavior of the spring operating mechanism. The fusion of multimodal data improves the depth and breadth of the analysis, making the results more reliable.

[0018] 5. After the stress sensors, velocity sensors and vibration sensors are arranged, the multi-frequency adaptive sampling algorithm is used to dynamically adjust the sampling frequency of key data according to the arrangement position of the sensors, which can ensure that data sampling is more intensive in key areas or key operating conditions, thereby capturing more detailed information. At the same time, in non-critical areas or stable operating conditions, the sampling frequency is appropriately reduced to reduce the burden of data storage and processing. This dynamic adjustment mechanism improves the flexibility and efficiency of data acquisition.

[0019] 6. Through multimodal data fusion and fault feature extraction, the operating status of the equipment can be more comprehensively reflected, and the accuracy and reliability of fault detection can be improved. The adaptive fault model can calculate the probability of fault occurrence in real time and give early warning before the fault occurs, which helps to reduce equipment downtime and improve production efficiency and economic benefits. Secondly, based on the importance assessment of fault features, a more scientific and reasonable maintenance strategy can be formulated, and key monitoring and maintenance of important fault features can be carried out to extend the service life of the equipment and reduce maintenance costs. The adaptive fault model can dynamically adjust parameters according to real-time data to adapt to changes in equipment status, enhance the adaptability and robustness of the system, and enable it to better cope with complex and changing operating environments.

[0020] 7. Automatic calibration and optimization can ensure that sensors can output accurate data under different working conditions, which helps reduce false alarms and missed alarms and improves the accuracy of fault detection. The reinforcement learning model can adaptively adjust sensor parameters to cope with changes in system status, enhance the robustness of the system, and enable it to better cope with complex and changing operating environments. Automatic calibration and optimization reduces the need for manual intervention and reduces maintenance costs. At the same time, due to the improved accuracy of sensor data, it can reduce unnecessary downtime and maintenance caused by false alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A flow chart of the spring operating mechanism fault detection and optimization method provided by the present invention; Figure 2 A geometric model diagram of the CT20 spring operating mechanism in the spring operating mechanism fault detection and optimization method provided by the present invention; Figure 3 A stress distribution cloud diagram in the spring operating mechanism fault detection and optimization method provided by the present invention; Figure 4 A time domain vibration signal diagram of the spring operating mechanism fault detection and optimization method provided by the present invention; Figure 5 This is a schematic diagram of the spring operating mechanism fault detection and optimization system provided by the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0024] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0025] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0026] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0027] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", which does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0028] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0029] In the field of fault detection of CT20 spring operating mechanism, it is of great significance to improve the accuracy and real-time performance of detection. Traditional fault detection methods usually rely on regular maintenance and manual inspection. However, these methods have many limitations and defects. For example, it is difficult for traditional methods to monitor the operating status of CT20 spring operating mechanism in real time, and it is easy to have the problem of false alarm or missed faults; secondly, when detecting the operating status of spring operating mechanism, only the data of individual components are collected through a single type of sensor, which makes the detection results not comprehensive enough and unable to accurately determine the type and level of faults. To this end, it is particularly important to improve the fault detection accuracy and adaptive detection ability of spring operating mechanism, realize real-time detection and fault prediction of CT20 spring operating mechanism, more comprehensively capture the operating status of equipment, and reduce false alarm and missed faults.

[0030] In order to solve the above technical defects, the inventor provides a spring operating mechanism fault detection and optimization method and related devices.

[0031] A first aspect of an embodiment of the present invention provides a spring operating mechanism fault detection and optimization method, such as Figure 1 As shown, including: S101, using stress sensors, velocity sensors and vibration sensors to collect key data of the spring operating mechanism in real time; illustratively, firstly construct a geometric model of the CT20 spring operating mechanism, such as Figure 2 As shown in the figure, the geometric model includes the main components of the spring operating mechanism, including springs, operating rods, connecting rod mechanisms and supporting structures. Each component needs to be accurately depicted in the geometric model to ensure accurate results in subsequent stress analysis. In order to perform accurate stress analysis, the material properties of each component in the spring operating mechanism geometric model are defined, including elastic modulus, yield strength and density; it is worth noting that the selection of material properties is usually based on the requirements of mechanical design and the actual working environment of the CT20 spring operating mechanism, as shown in the following table:

[0032] By defining the geometric model and the material properties of each component, the necessary basic data can be provided for the subsequent stress analysis. After the material properties of each component are defined, the material property data of each component in the above table are used to simulate the stress distribution and deformation behavior of the CT20 spring operating mechanism under the working load using finite element analysis (FEA), thereby providing an important reference for determining the installation position of the stress sensor.

[0033] Furthermore, according to the actual working state of the CT20 spring operating mechanism, the load conditions and boundary conditions are set. The load conditions include the force applied during the spring energy storage and release process, and the boundary conditions include fixed points and support points. Specifically, in the fully compressed state, a loading force of about 500 N is applied to the end of the spring, which acts along the axis of the spring to simulate the energy storage and release process of the spring. In addition, when the operating rod performs the switch operation, it needs to overcome an operating force of about 300 N, which is applied along the operating direction of the operating rod. In terms of boundary conditions, one end of the spring is fixed in the housing, so this end is set to be completely fixed, that is, the displacement and rotation of this point are constrained. At the same time, support boundary conditions are set at the connection points between the linkage mechanism and the support structure. The displacement of these points is restricted, but the rotational freedom is allowed to accurately simulate the actual working support situation.

[0034] Next, the geometric model of the spring operating mechanism is meshed, and the finite element analysis (FEA) tool is used to perform stress analysis. The stress distribution of the entire mechanism under the working load is calculated through FEA, and a preliminary stress distribution cloud map is generated, such as Figure 3 As shown; then, the adaptive stress sensing algorithm (ASSA) is used to analyze the stress distribution data and automatically identify the stress concentration area. According to the stress distribution cloud map, the data in the map is extracted into a matrix form, and the stress value of each grid point is As input data, calculate the stress gradient at each point:

[0035] in, is the stress value, and x, y, and z are the spatial coordinate axes.

[0036] The discrete form of stress gradient can be expressed as:

[0037]

[0038]

[0039] Through these formulas, the gradient value of each point in the stress distribution can be obtained and the area with significant stress change can be identified. Next, the threshold method is applied to set the threshold of the stress gradient. , when the stress gradient value ∇σ(x,y,z) at a point exceeds this threshold, the point is marked as a stress concentration point:

[0040] Then, the identified stress concentration points are clustered, and adjacent high stress gradient points are combined into a stress concentration area, which are potential sensor layout areas.

[0041] Subsequently, the ASSA algorithm was applied for multiple iterations to gradually adjust the stress gradient threshold. The stress concentration area is repeatedly calculated to ensure the maximum coverage of stress information in the area. At the same time, the ASSA algorithm is used to gradually refine the boundary of the stress concentration area, finally determine the part with the densest stress distribution, and lock the most suitable position for installing the stress sensor.

[0042] In the determined stress concentration area, the genetic optimization algorithm is applied to determine the optimal arrangement position of the sensor. The goal of this algorithm is to minimize the measurement redundancy of the sensor coverage area and maximize the capture of stress data. The objective function for optimizing the sensor position is:

[0043] in, is the stress value at the sensor arrangement point, is the average stress value in the area.

[0044] Finally, the layout plan of the stress sensors is generated, as shown in the following table, including the installation location, direction and precautions of each sensor, and the entire process and verification results are recorded to provide a basis for subsequent implementation.

[0045]

[0046] Before using the velocity sensor to collect the key data of the spring operating mechanism, it is also necessary to use the deployed strain sensors to obtain and record the initial motion data of the spring operating mechanism under different operating conditions. The initial motion data includes position, displacement, time and speed changes. According to the initial motion data, a motion model of the spring in the spring operating mechanism is established to simulate the dynamic behavior of the spring from static, motion and different operating conditions; then multimodal motion analysis is performed, using high-speed cameras and computer vision technology to capture the motion trajectory of the spring and its related components in real time, obtain visual data, and extract key motion features from these visual data, such as the displacement, velocity and acceleration of components such as springs and connecting rods in space. The specific steps include: Ensure that the visual data and the preliminary motion model data are synchronized in time and space. Typically, visual data (such as image sequences from high-speed cameras) and motion models (such as motion trajectories generated based on preliminary sensor data) may have different temporal sampling rates and spatial resolutions. Temporal alignment and spatial registration are required so that the two data can be compared and fused in the same coordinate system and timeline.

[0047] Then, key motion features (such as displacement, velocity, acceleration) are extracted from the visual data, and the motion vector field of the object is extracted using the optical flow method:

[0048] in, and are the gradients of the image in the x and y directions, is the rate of change of the image over time, u and v are the horizontal and vertical velocity components of the image point (i.e., motion vector). The corresponding motion parameters are extracted from the motion model, which are usually preliminarily estimated through sensor data, such as position change, velocity, acceleration, etc.

[0049] Using the established spring motion model, the motion parameters of the spring are extracted, and the Kalman filter is used to fuse the parameters from the visual data and the motion model. The Kalman filter is a recursive algorithm that can combine data from different sources while taking into account the noise level of each data source to generate an optimal estimate. The key formulas of the Kalman filter include: Status prediction:

[0050] This step is to predict the state of the system at the current time t (such as position, speed, etc.). is the predicted state vector, which is composed of the state vector at the previous moment It is obtained by transforming the state transfer matrix A, where B is the control input matrix, is the control input (such as external force or acceleration) applied at time t−1.

[0051] State prediction is used to estimate the future state (e.g., position, velocity) of a spring-actuated mechanism based on its previous state and control input (e.g., external force or operating action).

[0052] Error covariance prediction:

[0053] This step is the prediction error covariance matrix , the covariance matrix represents the uncertainty of the state estimate, the covariance matrix of the previous moment After the transformation of the state transfer matrix A and the process noise covariance matrix Q, the prediction error covariance matrix at the current moment is obtained , which can help decide where sensors should be placed in areas with larger errors to reduce uncertainty.

[0054] Kalman gain calculation:

[0055] Kalman Gain Is a weight matrix used to make the best combination between the predicted state and the actual observation, where is the observation matrix, which maps the state vector to the observation space, and R is the observation noise covariance matrix. The role of the Kalman gain is to measure the relative importance of the uncertainty of the prediction and the uncertainty of the observation. A larger gain means a more reliable observation, and a smaller gain means a more reliable prediction.

[0056] By calculating the Kalman gain, it is possible to determine how the contribution of visual data (actual observation) and motion model (prediction) should be weighed in the sensor placement. The size of the gain directly affects the estimation accuracy of the final motion characteristics, which in turn affects the optimization decision of sensor placement.

[0057] Status Update:

[0058] The state update formula is used to update the current state estimate by combining the predicted state and the actual observation. Here, is the observed value actually measured at time t, It is the predicted observation value obtained after mapping the predicted state to the observation space. The state update formula is calculated by Kalman gain. To adjust the predicted state to make it closer to the actual observation, thereby improving the accuracy of state estimation.

[0059] Error covariance update:

[0060] The error covariance update formula is used to update the covariance matrix at the current moment , which is the correction to the uncertainty in the state estimate after incorporating the observations. Since the observations introduce new information, the error covariance will usually decrease, meaning that the state estimate becomes more accurate.

[0061] The sensor data after Kalman filtering can reduce the estimation error of motion characteristics, thereby improving the accuracy of the motion model. By continuously updating the error covariance matrix, the sensor layout can be dynamically adjusted to adapt to the changing motion environment.

[0062] Next, MMMAA analysis is used to identify areas where the velocity of the spring and its related components changes significantly during movement. These areas are usually where stress or dynamic effects are concentrated and are most sensitive to velocity measurement. The specific steps include: Speed ​​change rate calculation: The speed change rate during the entire motion process is calculated through the fused motion model, that is, the derivative of speed with respect to time:

[0063] In the formula, v(t) represents the velocity at a certain time t, It represents the derivative of velocity with respect to time, that is, the instantaneous acceleration at time t Used to analyze the dynamic behavior of the CT20 spring actuator. By calculating the rate of change of velocity, it is possible to identify moments and areas in the system where the velocity changes significantly, which usually correspond to the acceleration or deceleration of the system.

[0064] Velocity gradient analysis: Calculate the gradient in the velocity field, that is, the rate of change of velocity in space. Areas with large velocity gradients mean that the velocity changes more dramatically, which is usually related to key events in the motion process. The velocity gradient calculation formula is:

[0065] In the formula, is the velocity gradient vector, which contains the rate of change of velocity in the x, y, and z directions, expressed as , , express.

[0066] The calculation of velocity gradient is used in this patent to identify areas where the velocity changes significantly during the movement of the CT20 spring operating mechanism. Areas with large velocity gradients usually mean that the velocity has changed dramatically in this area, which may be caused by critical events such as sudden spring release, stop or turn. These areas are usually where dynamic effects are concentrated, and it is suitable to install velocity sensors to ensure that these drastic changes are accurately captured.

[0067] The discrete form of the velocity gradient is:

[0068] The finite difference method is used to approximate the velocity gradient. The above formulas represent the discrete forms of the velocity gradient in the x, y, and z directions respectively.

[0069] For example, The discrete form of the calculation is the rate of change of velocity in the x direction, which is calculated by two adjacent points in the x direction. and The speed difference divided by the distance between them Come get.

[0070] Likewise, and The discrete form of is used to calculate the velocity gradients in the y and z directions respectively.

[0071] Determine the key points in the motion process where accurate speed measurement is most needed. These key points usually appear at the extreme positions of the motion trajectory, the area with the largest acceleration, or the area with high vibration frequency. This patent uses a genetic optimization algorithm to determine the layout of the speed sensor on the spring operating mechanism to ensure that the speed sensor can cover the most important motion area. The goal of this algorithm is to minimize the error of the speed sensor position and maximize the capture of key motion data. The objective function is set as:

[0072] in, is the velocity value of the sensor placement point, is the average speed value of the area.

[0073] Finally, based on the analysis results of MMMAA, the layout plan of the speed sensor is generated, as shown in the following table:

[0074] Similarly, before using the vibration sensor to collect the key data of the spring operating mechanism, it also includes using the existing vibration sensor or the temporarily arranged sensor to collect the vibration data of the spring operating mechanism under different operating conditions. These data include the time domain vibration signals of the spring under compression, release and other operating states, such as Figure 4 Then, the time domain vibration signal is converted to the frequency domain through Fourier transform (FFT):

[0075] in, represents the signal in the frequency domain, is the nth sampling point in the time domain, N is the total number of sampling points, k is the frequency index, and j is the imaginary unit.

[0076] And get the vibration amplitude :

[0077] in, and are the real and imaginary parts of the frequency domain signal respectively.

[0078] The vibration amplitude data is then combined with the previously collected stress and velocity data to further analyze the dynamic behavior of the spring operating mechanism under different operating conditions through multimodal fusion technology. The specific steps are: The vibration, stress, and velocity data were time synchronized and normalized to ensure that they were analyzed on the same time axis. For each data type, it was normalized to zero mean and unit variance:

[0079] in, , and represent vibration, stress and velocity data respectively, μ and σ are the mean and standard deviation of the corresponding data respectively.

[0080] The main features are extracted from the standardized data, and the following are obtained: Vibration characteristics :

[0081]

[0082] Among them, the peak represents the maximum amplitude of the vibration signal at time t, is the normalized vibration signal. Indicates the maximum value of the vibration signal in a given time period, the root mean square value Reflects the effective value of the vibration signal, and N is the total number of signal sampling points. represents the vibration signal value at the i-th sampling point, is the standardized vibration signal The Fourier transform result is used to obtain the features in the frequency domain. In the spectrum diagram, the frequency component with the highest amplitude is selected as the feature.

[0083] Stress characteristics :

[0084]

[0085] in, is the normalized stress signal. and Represent the maximum and minimum values ​​of the stress signal respectively. It is the gradient of the stress signal, which indicates the rate of change of stress in the spatial coordinate system.

[0086] Speed ​​characteristics :

[0087] in, The derivative with respect to time t. is the maximum value of the normalized velocity signal. is the standard deviation of the speed signal, which is used to indicate the breadth of speed distribution at different time points.

[0088] Fusion Vibration Features , stress characteristics and speed characteristics :

[0089] in, , , are weight coefficients corresponding to vibration, stress and velocity characteristics respectively, and these weights reflect the importance of each characteristic to the overall dynamic behavior. Fusion helps to determine the correlation between vibration and other motion characteristics (such as velocity and stress), thereby identifying the parts most prone to vibration.

[0090] Next, use MMMAA to identify the vibration mode of the spring in different operation modes. The specific steps are as follows: Extract the features related to the operation mode from the vibration, stress and speed features to form a feature vector:

[0091]

[0092]

[0093]

[0094]

[0095] in, represents the peak value of the vibration signal at time t, Indicates the maximum value of the vibration signal in a given time period; is the root mean square value (RMS) of the vibration signal, Represents the average value of the square value of the vibration signal of all sampling points; is the spectral characteristic of the vibration signal, which is obtained by Fourier transform Obtain, which represents the amplitude of the vibration signal in the frequency domain; is the maximum value of the stress signal, Represents the maximum value of the stress signal in a given time period; is the rate of change of the velocity signal, i.e., acceleration, given by It indicates the rate of change of speed over time. represents the normalized vibration signal, represents the normalized stress signal, Represents the normalized velocity signal.

[0096] K-means clustering is used to classify these features, group the operation modes, identify different vibration modes, and determine the main vibration characteristics by analyzing the frequency, amplitude, and duration T of each mode:

[0097]

[0098] where f is the value obtained by maximizing the spectrum amplitude to identify the main vibration frequencies, Representatives found that Maximum frequency; Indicates the maximum amplitude of the vibration signal, corresponding to the peak value of the vibration signal, and T is the signal duration, which is usually determined by calculating the time interval when the signal exceeds a certain threshold. Compare the vibration patterns under different operating modes to determine the operating conditions with the most severe vibration, and combine the stress and speed characteristics to determine the most vibration-sensitive parts. Based on the analysis results, mark high-amplitude or frequency-sensitive areas as candidate locations for sensor placement.

[0099] Finally, in order to further refine the position distribution of the sensors, a genetic optimization algorithm is used to optimize and determine the layout of the vibration sensors on the spring actuator. Through this algorithm, the goal is to minimize the measurement error of the vibration signal and maximize the coverage of the sensor on the key vibration data to ensure that the most critical vibration information can be captured in these sensitive areas:

[0100] in is the position of the i-th sensor, is the vibration acceleration value at this position, is the maximum vibration acceleration value in this area.

[0101] Finally, based on the analysis results of MMMAA, the layout plan of vibration sensors is generated, as shown in the following table:

[0102] When the spring operating mechanism is detected by the above method, the stress sensor, velocity sensor and vibration sensor closely related to the operation of the spring operating mechanism are used to collect the data of the spring operating mechanism in real time, so that the fault data can be quickly extracted, thereby realizing the rapid detection and warning of the spring operating mechanism fault; secondly, in the fault detection process, by combining three different types of collected data, the comprehensiveness and accuracy of the detection results are further improved, avoiding the problems of false alarms and missed alarms; finally, combined with the adaptive fault model, multimodal motion analysis algorithm and automatic calibration and optimization based on reinforcement learning, key data can be obtained and analyzed in real time during the operation of the equipment, greatly improving the timeliness of fault detection, and at the same time, the sensor can be optimized to ensure the reliability in the data acquisition process. By constructing an accurate geometric model of the spring operating mechanism and defining the material properties of each component in detail, a solid foundation is provided for the subsequent finite element analysis, ensuring the accuracy and reliability of the simulation results; at the same time, the stress distribution and deformation behavior of the spring operating mechanism under the working load are simulated by finite element analysis, which can intuitively display the stress state of the mechanism under different working conditions, help to discover potential stress concentration areas and deformation trends in advance, and provide an important reference for preventing faults. The load conditions and boundary conditions are set according to the actual working state of the spring operating mechanism, making the simulation results closer to the actual situation and improving the pertinence and practicality of the analysis; and the adaptive stress perception algorithm is used to deeply analyze the data on the stress distribution cloud map, automatically identifying the stress concentration area and the area with significant stress change, which not only improves the accuracy and efficiency of stress identification, but also reduces human intervention and errors, and provides a more accurate target area for the subsequent sensor layout. The spring motion model established based on the initial motion data can simulate the dynamic behavior of the spring from static, motion and different operating conditions, which helps to deeply understand the motion characteristics of the spring and provide theoretical support for subsequent data collection and analysis; the multimodal motion analysis technology can capture the motion trajectory of the spring in the spring operating mechanism in real time, and obtain rich visual data, extract key motion features, speed, acceleration, etc. from the visual data, and provide multi-dimensional information for data analysis. The Kalman filter is used to fuse the motion parameters extracted from the visual data and the spring motion model, which can effectively reduce data noise and errors and improve the accuracy and reliability of the data. The existing vibration sensors or temporarily arranged sensors are used to collect the vibration data of the spring operating mechanism under different operating conditions, which provides a rich data source for subsequent data analysis and makes the analysis results more comprehensive and accurate. Combining vibration data with stress data and velocity data and analyzing them through multimodal fusion technology can comprehensively consider the interactions and influences between different physical quantities, thereby more accurately revealing the dynamic behavior of the spring operating mechanism. The fusion of multimodal data improves the depth and breadth of the analysis, making the results more reliable.After the stress sensors, velocity sensors, and vibration sensors are arranged, the multi-frequency adaptive sampling algorithm is used to dynamically adjust the sampling frequency of key data according to the arrangement position of the sensors, which can ensure that data sampling is more intensive in key areas or key operating conditions, thereby capturing more detailed information. At the same time, in non-critical areas or stable operating conditions, the sampling frequency is appropriately reduced to reduce the burden of data storage and processing. This dynamic adjustment mechanism improves the flexibility and efficiency of data acquisition.

[0103] S102. Based on the collected key data, a multidimensional data framework is constructed to obtain a multidimensional data set. Exemplarily, after the stress sensors, velocity sensors, and vibration sensors are arranged, a multi-frequency adaptive sampling algorithm is used to dynamically adjust the sampling frequency of the key data according to the arrangement positions of the stress sensors, velocity sensors, and vibration sensors, so that the stress, velocity, and vibration multidimensional data are collected simultaneously.

[0104] First, the multi-frequency adaptive sampling algorithm (MFASA) dynamically adjusts the sampling frequency of each sensor based on its location and the characteristics it monitors. Specifically, for sensors that detect critical state changes, MFASA increases the sampling frequency to ensure high-resolution data acquisition. The sampling frequency of each sensor at time tt is It can be expressed as:

[0105] in, is the initial sampling frequency, is the adjustment coefficient, is the sensor output signal at time t The rate of change, is the maximum rate of change of the sensor output signal.

[0106] Then, the synchronous data acquisition system is used at each sampling time. Collect data from all sensors simultaneously, ensuring that stress, velocity, and vibration data are collected at the same time. Stress Data , speed data and vibration acceleration data Obtained from stress sensor, velocity sensor and vibration sensor respectively. Each data value can be expressed as:

[0107]

[0108]

[0109] in, , and Respectively indicate at time Stress, velocity, and vibration acceleration data at the (x, y, z) coordinates.

[0110] Finally, multi-dimensional data synchronization processing uses time alignment technology to ensure that data collected at different frequencies can be compared and analyzed on the same time axis. Organized into a unified data structure for easy subsequent storage and processing:

[0111] In this data structure, Represents the comprehensive data collected at time t, including stress ,speed and vibration acceleration This step ensures the coordination and consistency between the data, providing high-quality data input for subsequent analysis and modeling.

[0112] A multidimensional data framework is constructed on the stress, velocity and vibration multidimensional data collected simultaneously, and the collected stress, velocity and vibration multidimensional data are standardized and integrated by using multimodal data fusion technology; Data from different sensors are processed and stored synchronously to ensure the compatibility and coordination between different types of data, providing complete and accurate data input for subsequent data analysis and prediction.

[0113] First, various data are time-aligned and standardized to ensure that all data are analyzed synchronously on the same time axis. For time point, is the stress data, is the speed data, is the vibration data, is the temperature data. These data are standardized to zero mean and unit variance to eliminate the impact of dimensional differences:

[0114] in, , , , are the means of stress, velocity, vibration and temperature data respectively, , , , is its standard deviation.

[0115] Then, the standardized data is integrated into a unified three-dimensional data matrix, where the first dimension is time, the second dimension is sensor location, and the third dimension is data type. Location( ) can be represented as:

[0116] in, It's in time and location ( ) contains normalized stress, velocity, vibration, and temperature data.

[0117] Finally, a multidimensional data table is generated to store the data in a structured manner for subsequent analysis and processing. This table can be regarded as a record in the database, each record contains time, location coordinates and various data values. The final table format is as follows:

[0118] S103, extracting fault features and performing data analysis from the obtained multidimensional data set, and constructing an adaptive fault model based on the extracted fault features, and then obtaining real-time fault detection results through multimodal data fusion; illustratively, extracting stress data, velocity data, and vibration data fault features from the multidimensional data set. It is worth noting that the data is not limited to the above three types. By extracting these features, the most valuable information for fault prediction can be extracted from the complex data set, specifically including: Stress feature extraction: By analyzing stress data, the stress peak value is extracted and stress gradient The stress peak characteristic indicates the maximum stress that the system is subjected to during operation, and the stress gradient indicates the rate of change of stress at different locations:

[0119] in, represents the stress value changing with time, Represents the rate of change of stress in each spatial direction.

[0120] Speed ​​feature extraction: Analyze speed data and extract the speed change rate and peak speed The speed change rate can identify the dynamic response characteristics of the system, and the speed peak value indicates the maximum speed reached by the system during operation:

[0121] in, is the speed data, Indicates the rate of change of velocity with time (acceleration).

[0122] Vibration Feature Extraction: Extracting RMS values ​​from vibration signals and frequency domain characteristics . RMS value of vibration Usually used to evaluate the vibration intensity of the system, frequency domain characteristics The frequency content of a signal can be analyzed:

[0123] in, is the vibration data, Indicates the effective value of the signal, It is the frequency domain feature of the vibration signal, which is calculated by Fourier transform (FFT).

[0124] Various features such as stress, speed, vibration, etc. are extracted from the multidimensional data framework, and these features are further screened and optimized to identify the most critical features for fault prediction, thereby improving the accuracy and efficiency of the prediction model. First, a correlation analysis is performed to determine which features are most closely related to the occurrence of faults. The main task of this step is to identify those features that are highly correlated with the health status of the system, especially the signs of faults. The specific steps include: Calculate Pearson's correlation coefficient: Using Pearson's correlation coefficient To measure the linear correlation between different fault feature data.

[0125]

[0126] in represents the correlation coefficient between features x and y, and are their means, respectively. Features with values ​​close to 1 or -1 indicate strong correlation, while values ​​close to 0 indicate weak correlation.

[0127] The correlation calculation structure corresponding to all fault features is formed into a correlation matrix to identify fault features with mutual correlation and fault features containing redundant information; then, after completing the correlation analysis, the importance of each feature in fault prediction is further evaluated. This step is mainly completed through the feature importance algorithm to ensure that the features with the most predictive value are retained: The importance of each fault feature in fault prediction is evaluated using the random forest algorithm:

[0128] in, Representation characteristics The importance of It is a feature Gini index contribution of the t-th tree, where T is the total number of trees in the random forest.

[0129] Principal Component Analysis (PCA): Use PCA to perform dimensionality reduction analysis, project high-dimensional feature data into a low-dimensional space, and extract principal components, which reflect the main change patterns of the data.

[0130]

[0131] Among them, Z is the data matrix after dimensionality reduction, X is the original feature matrix, and W is the weight vector matrix of the feature matrix. PCA identifies the most important feature combination by maximizing the variance of the projected data.

[0132] Finally, based on the feature importance evaluation results, the most important features are selected and further optimized to enhance the effect of fault prediction: Remove redundant features: Based on correlation analysis and feature importance scores, remove redundant features that are highly correlated with other features to reduce the complexity of the model and the risk of overfitting.

[0133] Optimize feature set: Combine the results of feature importance and principal component analysis to optimize the final feature set to ensure that only the features with the most predictive value are retained. The optimized feature set can not only improve the predictive performance of the model but also reduce the computational cost.

[0134] Based on the extracted fault features, an adaptive fault model is established, and the adaptive fault model is used to calculate the probability of fault occurrence in real time. By combining Bayesian networks and fuzzy logic technology, AFPM can accurately assess fault risks in complex and uncertain environments and provide reliable probability outputs for subsequent intelligent warnings. The core task of AFPM is to calculate the probability of system failure by processing the extracted and optimized key features in real time. These key features may include parameters such as stress, speed, and vibration. AFPM performs probabilistic reasoning through Bayesian networks and combines fuzzy logic to process the ambiguity of input data, and finally derives the probability value of fault occurrence.

[0135] First, Bayesian network modeling is performed. Bayesian network is the core component of AFPM and is used to infer the probability of system failure based on existing data: Node definition: In the Bayesian network, the key features extracted by S3 (such as stress peak , speed change rate (t) and vibration RMS value etc.) are regarded as input nodes, and fault states are regarded as output nodes.

[0136] Construction of Conditional Probability Tables (CPT): Using historical data, conditional probability tables (CPT) are constructed. These tables show the probability of failure under different characteristic conditions. For example, given a stress peak When a certain threshold is exceeded, the probability of the system entering a fault state can be obtained through statistical data.

[0137] Bayesian reasoning: Calculate the posterior probability of a fault through the Bayesian formula

[0138] in, is the probability of observing these characteristic data under a given fault state, is the prior probability of failure occurrence, is the total probability of observing these data.

[0139] Then, the fuzzy logic system is integrated. In order to deal with the uncertainty of input data, AFPM combines the fuzzy logic system to process fuzzy data and enhance the accuracy of fault probability assessment: Fuzzy variable definition: Convert key features into fuzzy variables. For example, convert the peak stress Convert to fuzzy sets (such as "low", "medium", "high") and set the fuzzy membership function according to the distribution of actual data.

[0140] Fuzzy rule base: Build a fuzzy rule base to define the failure probability under different fuzzy states. For example, the rule can be set as “if is 'high', and is 'high', then the probability of failure is 'high'".

[0141] Fuzzy reasoning: Fuzzy variables are mapped to failure probabilities through the fuzzy reasoning mechanism. The output of fuzzy reasoning is combined with the probability results of the Bayesian network to obtain a more accurate estimate of the failure probability.

[0142] Finally, AFPM calculates the comprehensive probability of system failure by integrating the reasoning results of the Bayesian network and the fuzzy logic system:

[0143] in, is the weight coefficient, the value range is 0≤ ≤10, which is used to balance the influence of Bayesian network and fuzzy logic system on the final result. is the failure probability derived from the Bayesian network inference. is the failure probability derived by the fuzzy logic system.

[0144] This failure probability value It is an important input for subsequent decision-making and early warning systems and will be used to generate specific warning information and maintenance recommendations in subsequent steps.

[0145] Through multimodal data fusion technology, stress data, velocity data and vibration data are comprehensively analyzed to obtain fault detection results and warning information, and the warning level is adaptively adjusted according to different fault types, providing key reference information for system maintenance and optimization.

[0146] First, the failure probability calculated by AFPM The probability is passed to the intelligent warning algorithm (MDLSWA). MDLSWA analyzes this probability and determines whether it is necessary to generate a warning message. The judgment standard is usually that the probability of failure exceeds a certain preset threshold. :

[0147] When the probability of failure exceeds the threshold, MDLSWA generates preliminary warning information, identifying the type and severity of the possible failure. Then, multimodal data fusion, in order to improve the accuracy and real-time nature of the warning information, MDLSWA uses cross-modal data fusion technology to comprehensively analyze multimodal data (such as stress, speed, vibration, etc.) from different sensors. This process includes: Data synchronization and fusion: MDLSWA uses cross-modal data fusion technology to synchronize data of different modes (stress, speed, vibration, etc.) in time and space, and conducts comprehensive processing based on their respective feature weights.

[0148] Deep learning analysis: The fused data is fed into a deep learning model that has been trained to recognize the characteristic patterns in different failure modes. The model analyzes the fused data to generate more accurate failure predictions and warning information.

[0149] Next, adaptive warning level adjustment, MDLSWA not only generates basic warning information, but also adaptively adjusts the warning level according to the fault type and severity: Warning level setting: Based on the fault type, possible impact scope and current system status output by the deep learning model, MDLSWA divides warning information into different levels (such as "low", "medium" and "high").

[0150] Real-time adjustment: During system operation, MDLSWA dynamically adjusts the alert level based on real-time data changes. If the system status deteriorates or a more serious failure mode is detected, the alert level will automatically increase.

[0151] Finally, the intelligent warning information generated by MDLSWA includes: Fault type identification: Point out the specific fault type that may occur (such as stress fatigue, abnormal vibration, etc.).

[0152] Failure probability: Displays the comprehensive probability of failure occurrence .

[0153] Warning level: indicates the severity of the warning (such as "low risk", "medium risk", "high risk").

[0154] S104. By using the real-time fault detection results, the stress sensor, velocity sensor and vibration sensor are optimized by using automatic calibration and optimization of reinforcement learning. For example, according to the real-time fault detection results and warning information, the stress sensor, velocity sensor and vibration sensor are automatically calibrated and optimized by using automatic calibration and optimization of reinforcement learning.

[0155] Based on the warning information provided by MDLSWA, reinforcement learning-based automatic calibration and optimization (RLABO) automatically calibrates the sensor to ensure its accuracy in future operations. RLABO further optimizes the calibration strategy through reinforcement learning to ensure data collection and analysis for the next operation cycle, thereby enhancing the long-term stable operation of the CT20 spring actuator.

[0156] After MDLSWA generates an alert, the sensor calibration process is automatically triggered based on the alert level and type. RLABO gradually adjusts the sensor calibration parameters in the operating environment through real-time learning. The state space of the reinforcement learning model includes the current calibration state of the sensor, environmental conditions, and historical data, while the action space involves adjustable calibration parameters such as the zero offset and sensitivity adjustment of the sensor. The effect of the calibration is measured by the reward function, which is It is determined by the accuracy of the sensor data after calibration and the accuracy of the system detection. The formula is:

[0157] in, is the calibrated sensor data, is the ideal sensor data, is the cost of performing the calibration action a, and λ is a coefficient that weighs accuracy and cost.

[0158] RLABO automatically adjusts sensor parameters based on the decision results of reinforcement learning to ensure its high accuracy in future operations, and records the updated parameters to provide a reference for the next cycle of operation. As data is continuously collected, RLABO will continue to optimize the calibration strategy so that the system can maintain the best state under different operating conditions.

[0159] Closed-loop intelligent optimization: After maintenance is completed, the closed-loop intelligent optimization algorithm (CLIOA) module will start after completing the automatic calibration of the sensor to globally optimize the entire fault detection system. CLIOA automatically adjusts algorithm parameters and detection processes by reviewing historical data and predicting future data. First, a genetic algorithm is applied. Each individual represents a parameter combination of a fault detection model, such as feature weights and model hyperparameters. The fitness function is determined by the model's fault detection accuracy and operating efficiency. The formula is:

[0160] Where x represents the parameter set of the model, is a coefficient that weighs accuracy and computational cost. Then, CLIOA uses the particle swarm optimization (PSO) algorithm to further adjust the model parameters. Each particle represents a possible parameter setting, and the update rule is:

[0161] in, is the velocity of particle i, is the position of the particle, is the individual optimal position of the particle, g is the global optimal position, ω, , is the control parameter, , is a random number. CLIOA combines genetic algorithm and particle swarm optimization to perform multiple rounds of iterations to find the best parameter settings to ensure that the fault detection system can maintain the most efficient operation under different working conditions.

[0162] Finally, CLIOA updates all system parameters and saves the optimization results, ensuring that the entire system can work with maximum efficiency and precision during future operation cycles.

[0163] A second aspect of the embodiments of the present invention provides a spring operating mechanism fault detection and optimization system, such as Figure 5 As shown, including: The acquisition module is used to collect key data of the spring operating mechanism in real time; A framework building module, used to build a multidimensional data framework based on the collected key data to obtain a multidimensional data set; A fault feature extraction module, used for extracting fault features from the obtained multi-dimensional data set; Data analysis module, used for data analysis of multidimensional data sets; A model building module for building an adaptive fault model; Reinforcement learning module for optimizing strain, velocity, and vibration sensors.

[0164] According to a third aspect of an embodiment of the present invention, there is provided an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable in the processor, wherein the processor implements the steps of the spring operating mechanism fault detection and optimization method as described above when executing the computer program.

[0165] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the spring operating mechanism fault detection and optimization method as described above is implemented.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims of the invention.

Claims

1. A spring operating mechanism fault detection and optimization method, characterized in that: include: Use stress sensors, velocity sensors and vibration sensors to collect key data of the spring operating mechanism in real time; Based on the collected key data, a multidimensional data framework is constructed to obtain a multidimensional data set; Fault feature extraction and data analysis are performed from the obtained multidimensional data set, and an adaptive fault model is constructed based on the extracted fault features, and then a real-time fault detection result is obtained through multimodal data fusion; The stress sensor, velocity sensor and vibration sensor are optimized through the real-time fault detection results and automatic calibration and optimization using reinforcement learning.

2. The fault detection and optimization method according to claim 1, characterized in that: Before the key data of the spring operating mechanism is collected by the stress sensor, the method further includes: Constructing a geometric model of a spring operating mechanism, defining material properties of various components in the geometric model of the spring operating mechanism, and simulating stress distribution and deformation behavior of the spring operating mechanism under working load by using finite element analysis; According to the actual working state of the spring operating mechanism, the load conditions and boundary conditions are set, and the geometric model of the spring operating mechanism is meshed to generate a preliminary stress distribution cloud map; Using an adaptive stress perception algorithm, the stress distribution data on the stress distribution cloud map is analyzed to automatically identify stress concentration areas and obtain the gradient value of each point in the stress distribution and the area with significant stress changes; The identified stress concentration points are clustered so that adjacent high stress gradient points are combined into a stress concentration area; The ASSA algorithm is used to refine the boundary of the stress concentration area and determine the intensive stress distribution area. The genetic optimization algorithm is used to determine the layout position of the stress sensor on the spring operating mechanism.

3. The fault detection and optimization method according to claim 1, characterized in that: Before the key data of the spring operating mechanism is collected by the speed sensor, the method further includes: Using the arranged strain sensors, the initial motion data of the spring operating mechanism under different operating states are obtained, and based on the initial motion data, a motion model of the spring in the spring operating mechanism is established to simulate the dynamic behavior of the spring under static, moving and different operating conditions; Using multimodal motion analysis to capture the motion trajectory of the spring in the spring operating mechanism in real time, obtain visual data, and extract key motion features from the visual data; Utilizing the established spring motion model, extracting the motion parameters of the spring, and using a Kalman filter to fuse the visual data and the motion parameters; The arrangement position of the speed sensor on the spring operating mechanism is determined by a genetic optimization algorithm.

4. The fault detection and optimization method according to claim 1, characterized in that: Before the key data of the spring operating mechanism is collected by the vibration sensor, the method further includes: Use existing vibration sensors or temporarily deployed sensors to collect vibration data of the spring operating mechanism under different operating conditions; Combining vibration data with stress data and velocity data, and using multimodal fusion technology, the dynamic behavior of the spring operating mechanism under different operating conditions is analyzed; Using multimodal motion analysis algorithms, the vibration modes of the spring in different operating modes are identified; The vibration patterns under different operating modes are compared to determine the operating conditions with the most severe vibration, and the layout position of the vibration sensor on the spring operating mechanism is determined by a genetic optimization algorithm.

5. The fault detection and optimization method according to claim 1, characterized in that: Based on the collected key data, a multidimensional data framework is constructed to obtain a multidimensional data set, which specifically includes: After the stress sensor, velocity sensor and vibration sensor are arranged, a multi-frequency adaptive sampling algorithm is used to dynamically adjust the sampling frequency of key data according to the arrangement positions of the stress sensor, velocity sensor and vibration sensor, so that stress, velocity and vibration multi-dimensional data can be collected simultaneously; A multidimensional data framework is constructed on the stress, velocity and vibration multidimensional data collected simultaneously, and the collected stress, velocity and vibration multidimensional data are standardized and integrated by using multimodal data fusion technology; Integrate the standardized multi-dimensional data into a three-dimensional data matrix, wherein the first dimension is time, the second dimension is sensor position, and the third dimension is data type; Based on the three-dimensional data matrix, a multidimensional data table is generated to obtain a multidimensional data set.

6. The fault detection and optimization method according to claim 1, characterized in that: Fault feature extraction and data analysis are performed from the obtained multidimensional data set, and an adaptive fault model is constructed based on the extracted fault features. Then, a real-time fault detection result is obtained through multimodal data fusion, specifically including: Extract stress data, velocity data and vibration data fault features from the multidimensional data set, and use the Pearson correlation coefficient to measure the linear correlation between different fault feature data; The correlation calculation structures corresponding to all fault features are combined into a correlation matrix to identify fault features with mutual correlation and fault features containing redundant information; The random forest algorithm is used to evaluate the importance of each fault feature in fault prediction, and the most important fault features are selected based on the fault feature importance evaluation structure to optimize the fault prediction effect; Based on the extracted fault features, an adaptive fault model is established, and the probability of fault occurrence is calculated in real time using the adaptive fault model; Through multimodal data fusion technology, stress data, velocity data and vibration data are comprehensively analyzed to obtain fault detection results and warning information.

7. The fault detection and optimization method according to claim 1, characterized in that: Through the real-time fault detection results, the stress sensor, velocity sensor and vibration sensor are optimized by using automatic calibration and optimization of reinforcement learning, specifically including: Based on real-time fault detection results and warning information, automatic calibration and optimization of stress sensors, velocity sensors, and vibration sensors are automatically optimized using reinforcement learning.

8. A spring operating mechanism fault detection and optimization system, characterized in that: include: The acquisition module is used to collect key data of the spring operating mechanism in real time; A framework building module, used to build a multidimensional data framework based on the collected key data to obtain a multidimensional data set; A fault feature extraction module, used for extracting fault features from the obtained multi-dimensional data set; Data analysis module, used for data analysis of multidimensional data sets; A model building module for building an adaptive fault model; Reinforcement learning module for optimizing strain, velocity, and vibration sensors.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable in the processor, wherein the processor implements the steps of the spring operating mechanism fault detection and optimization method according to any one of claims 1 to 7 when executing the computer program.

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 spring operating mechanism fault detection and optimization method according to any one of claims 1 to 7 is implemented.

Images