A rail transit nondestructive testing device and a testing method thereof

Abstract

The present invention provides a rail transit nondestructive testing device and a testing method thereof, the method comprising: acquiring testing data for rail transit equipment; performing data preprocessing on the testing data to obtain preprocessed data, and extracting feature vectors based on the preprocessed data; based on the feature vectors and the preprocessed data, performing data enhancement on the testing data through the few-step diffusion model to obtain synthetic data; based on the testing model, constructing a jump data diffusion model, performing robustness analysis on the testing data based on the jump data diffusion model to obtain robustness analysis results; based on the robustness analysis results, using a preset implicit neural representation network to extract time continuity features in the testing data, and performing time series anomaly detection on the time series anomaly detection to obtain time series anomaly detection results.

Description

Technical Field

[0001] The present application relates to the field of rail transit technology, and in particular to a rail transit nondestructive testing device and a testing method thereof. Background Art

[0002] As an important mode of transportation, the safety and reliability of rail transit are directly related to the safety of people's lives and property. Existing rail transit equipment detection technologies mainly include ultrasonic detection, temperature monitoring and vibration monitoring. Among them, ultrasonic detection can detect defects inside the equipment, temperature monitoring can timely detect abnormal heating during the operation of the equipment, and vibration monitoring can detect abnormal vibration during the operation of the equipment.

[0003] However, the existing technology has the following problems: first, the reliability and accuracy of a single detection method are limited, and it is difficult to fully reflect the operating status of the equipment; second, the fusion analysis capability of multi-source detection data is insufficient, and it is impossible to fully utilize the data collected by various sensors.

[0004] Therefore, the existing rail transit equipment detection technology has certain limitations and low detection accuracy. Summary of the invention

[0005] In view of this, the present application provides a rail transit non-destructive testing device and a testing method thereof, which solves the problem that the rail transit equipment testing technology in the prior art has limitations and low detection accuracy.

[0006] In a first aspect, an embodiment of the present application provides a rail transit nondestructive testing method, comprising:

[0007] Acquire detection data for rail transit equipment, the detection data including ultrasonic reflection signals from a multi-channel ultrasonic array detection system, and temperature data and vibration data from a multi-source sensor network system;

[0008] Performing data preprocessing on the detection data to obtain preprocessed data, and extracting a feature vector based on the preprocessed data;

[0009] Based on the feature vector and the preprocessed data, a few-step diffusion model is constructed, and data enhancement is performed on the detection data through the few-step diffusion model to obtain synthetic data;

[0010] Based on the synthetic data, a dense reward difference learning framework is constructed, and the dense reward difference learning framework is optimized to obtain a detection model;

[0011] Based on the detection model, a jump data diffusion model is constructed, and based on the jump data diffusion model, a robustness analysis is performed on the detection data to obtain a robustness analysis result;

[0012] Based on the robustness analysis result, a preset implicit neural representation network is used to extract the time continuity feature in the detection data, and a time series anomaly detection is performed on the time series anomaly feature to obtain a time series anomaly detection result;

[0013] Multi-model data fusion is performed on the abnormal data detection result, the robustness analysis result and the time series anomaly detection result to obtain a detection warning result, wherein the abnormal data detection result is obtained after the detection model performs anomaly detection on the detection data.

[0014] In a possible implementation, the preprocessing the detection data to obtain preprocessed data includes:

[0015] Performing wavelet transform processing on the ultrasonic reflection signal, processing the high-frequency coefficients obtained after the wavelet transform, and reconstructing the signal based on the processed high-frequency coefficients;

[0016] Performing median filtering on the temperature data to eliminate mutation points in the temperature data, and performing low-pass filtering on the temperature data to eliminate high-frequency noise in the temperature data;

[0017] Performing bandpass filtering on the vibration data and processing abnormal value points in the vibration data;

[0018] Based on the timestamp information, the processed ultrasonic reflection signal, the processed temperature data and the processed vibration data are aligned and standardized to obtain preprocessed data.

[0019] In a possible implementation, the preprocessed data includes a preprocessed ultrasonic reflection signal, preprocessed temperature data, and preprocessed vibration data, and extracting a feature vector based on the preprocessed data includes:

[0020] Performing time domain feature extraction and frequency domain feature extraction on the preprocessed ultrasonic reflection signal to obtain ultrasonic reflection signal feature data;

[0021] Extracting temperature characteristic data of the preprocessed temperature data, wherein the temperature characteristic data includes temperature gradient, temperature field distribution characteristics and temperature fluctuation characteristics;

[0022] Extracting vibration feature data of the preprocessed vibration data, wherein the vibration feature data includes time domain statistical features, frequency domain features, and time domain joint features;

[0023] The ultrasonic reflection signal characteristic data, the temperature characteristic data and the vibration characteristic data are screened and evaluated to obtain a characteristic vector.

[0024] In a possible implementation, constructing a dense reward difference learning framework based on the synthetic data, optimizing the dense reward difference learning framework, and obtaining a detection model includes:

[0025] Based on the synthetic data, construct a local feature reward function and a global feature reward function, wherein the local feature reward function is used to indicate a reward for the similarity of local features, and the global feature reward function is used to indicate a reward for the consistency of global features;

[0026] Based on the local feature reward function and the global feature reward function, construct a dense reward difference learning framework;

[0027] Based on the dense reward difference learning framework, calculating sample weights, and performing difference optimization based on contrastive learning of the synthetic data;

[0028] Optimizing parameters of the dense reward difference learning framework based on the sample weights;

[0029] The dense reward difference learning framework with optimized parameters is evaluated and optimized to obtain the detection model.

[0030] In a possible implementation, constructing a jump data diffusion model based on the detection model, performing robustness analysis on the detection data based on the jump data diffusion model, and obtaining a robustness analysis result includes:

[0031] Based on the detection model, a multi-scale feature extraction network is constructed, wherein the multi-scale feature extraction network includes a skip connection structure in which a plurality of feature scale layers arranged in cascade are arranged between adjacent feature scale layers;

[0032] Constructing a feature fusion mechanism, wherein the feature fusion mechanism is used to indicate that a preset channel attention mechanism and a spatial attention mechanism are used to perform feature fusion on the detection data;

[0033] Constructing an anti-interference network, wherein the anti-interference network is used to instruct to perform feature alignment on the detection data using a preset spatial transformation network;

[0034] Based on the multi-scale feature extraction network, the feature fusion mechanism and the anti-interference network, a jump data diffusion model is constructed;

[0035] Performing system performance evaluation and optimization on the jump data diffusion model;

[0036] Based on the optimized jump data diffusion model, a robustness analysis is performed on the detection data to obtain a robustness analysis result.

[0037] In a possible implementation, based on the robustness analysis result, using a preset implicit neural representation network to extract the time continuity feature in the detection data, and performing time series anomaly detection on the time series anomaly feature to obtain the time series anomaly detection result includes:

[0038] Based on the robustness analysis result, constructing an implicit neural representation network, wherein the implicit neural representation network includes a time continuity encoder, a temporal dependency model and an anomaly detection discriminator;

[0039] Extracting a temporal continuity feature in the detection data based on the implicit neural representation network, and calculating an abnormality score of the temporal continuity feature;

[0040] The anomaly score of the time continuity feature is judged based on a preset anomaly threshold to obtain a corresponding time series anomaly detection result.

[0041] In a possible implementation, performing multi-model data fusion on the abnormal data detection result, the robustness analysis result, and the time series abnormality detection result to obtain the detection warning result includes:

[0042] Assigning corresponding weight coefficients to the abnormal data detection result, the robustness analysis result and the timing anomaly detection result;

[0043] Performing weighted fusion on the abnormal data detection result, the robustness analysis result and the time series abnormality detection result according to the weight coefficient;

[0044] Output the corresponding detection and warning results based on the weighted fusion results.

[0045] In a possible implementation, before acquiring the detection data for the rail transit equipment, the method further includes:

[0046] Obtaining ultrasonic testing requirements, environmental monitoring requirements and structural characteristics of the rail transit equipment;

[0047] Based on the ultrasonic detection requirements and the structural characteristics of the rail transit equipment, determine the structural strategy and ultrasonic excitation parameters of the multi-channel ultrasonic array, wherein the structural strategy includes the number, spacing and arrangement of array elements;

[0048] Configuring ultrasonic signal acquisition parameters of the multi-channel ultrasonic array, wherein the ultrasonic signal acquisition parameters include ultrasonic signal sampling frequency and quantization accuracy;

[0049] Based on the structural strategy of the multi-channel ultrasonic array, the ultrasonic signal acquisition parameters and the ultrasonic excitation parameters, an ultrasonic array arrangement strategy is formed and sent to the multi-channel ultrasonic array detection system, so that the multi-channel ultrasonic array detection system is arranged based on the ultrasonic array arrangement strategy;

[0050] Based on the environmental monitoring requirements and the structural characteristics of the rail transit equipment, determine the layout strategy of the temperature sensor array and the deployment strategy of the vibration sensor;

[0051] Configuring sensor data acquisition parameters of the temperature sensor array and the vibration sensor, wherein the sensor data acquisition parameters include sensor data sampling frequency and triggering mode;

[0052] Based on the layout strategy of the temperature sensor array, the deployment strategy of the vibration sensor and the sensor data acquisition parameters, a sensor network deployment strategy is formed and sent to the multi-source sensor network system, so that the multi-source sensor network system is deployed based on the sensor network deployment strategy.

[0053] In a possible implementation, constructing a few-step diffusion model based on the feature vector and the preprocessed data includes:

[0054] Based on the feature vector, construct a few-step diffusion initial model, wherein the few-step diffusion initial model includes a basic model architecture, a noise scheduling strategy, a sampling mechanism, a loss function, and an optimization strategy;

[0055] Dividing the feature vector and the preprocessed data into a training set, a validation set and a test set;

[0056] Using the training set to train the few-step diffusion initial model;

[0057] Using the validation set to evaluate and optimize the trained initial model of few-step diffusion;

[0058] The optimized few-step diffusion initial model is tested using the test set to obtain the few-step diffusion model.

[0059] In a second aspect, the present application also provides a rail transit nondestructive testing device, including:

[0060] processor;

[0061] memory for storing computer programs;

[0062] Wherein, when the processor executes the computer program, the steps of the rail transit non-destructive testing method as described in the above-mentioned first aspect or any possible implementation method of the first aspect are implemented.

[0063] The rail transit nondestructive testing method and equipment provided by the present application solves the problem of a single detection method by acquiring ultrasonic reflection signals from a multi-channel ultrasonic array detection system for rail transit equipment, and temperature data and vibration data from a multi-source sensor network system for rail transit equipment, thereby realizing the collection of multi-source detection data and improving the comprehensiveness and accuracy of the detection. Subsequently, by preprocessing and extracting features of the multi-source detection data and using a few-step diffusion model for data enhancement, the fusion capability of the multi-source detection data is improved, and the training requirements of subsequent models can be met. Furthermore, by constructing a few-step diffusion model and a dense reward difference learning framework, the accuracy of abnormal data detection is improved. By constructing a jump data diffusion model, the robustness of the detection results of the detection model is improved, and the detection model can be applied to different working conditions and environmental conditions. The time continuity features are extracted through an implicit neural representation network for time series anomaly detection, and a deep analysis of the time series features of the detection data is realized, thereby performing multi-model data fusion on the abnormal data detection results, the robustness analysis results, and the time series anomaly detection results, thereby improving the detection accuracy and the reliability of the detection warning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the disclosed embodiments in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Figure 1 It is a flow chart of a rail transit nondestructive testing method provided in an embodiment of the present application;

[0066] Figure 2 A schematic diagram of the structure of a rail transit nondestructive testing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0067] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0068] Figure 1 It is a flow chart of a rail transit nondestructive testing method provided in an embodiment of the present application. Based on the example flow, the embodiment of the present application provides a rail transit nondestructive testing method, including the following steps S101 to S107.

[0069] S101. Acquire detection data for rail transit equipment, wherein the detection data includes ultrasonic reflection signals from a multi-channel ultrasonic array detection system, and temperature data and vibration data from a multi-source sensor network system.

[0070] In some embodiments, before step S101, the method further includes:

[0071] Obtaining ultrasonic testing requirements, environmental monitoring requirements and structural characteristics of the rail transit equipment;

[0072] Based on the ultrasonic detection requirements and the structural characteristics of the rail transit equipment, determine the structural strategy and ultrasonic excitation parameters of the multi-channel ultrasonic array, wherein the structural strategy includes the number, spacing and arrangement of array elements;

[0073] Configuring ultrasonic signal acquisition parameters of the multi-channel ultrasonic array, wherein the ultrasonic signal acquisition parameters include ultrasonic signal sampling frequency and quantization accuracy;

[0074] Based on the structural strategy of the multi-channel ultrasonic array, the ultrasonic signal acquisition parameters and the ultrasonic excitation parameters, an ultrasonic array arrangement strategy is formed and sent to the multi-channel ultrasonic array detection system, so that the multi-channel ultrasonic array detection system is arranged based on the ultrasonic array arrangement strategy;

[0075] Based on the environmental monitoring requirements and the structural characteristics of the rail transit equipment, determine the layout strategy of the temperature sensor array and the deployment strategy of the vibration sensor;

[0076] Configuring sensor data acquisition parameters of the temperature sensor array and the vibration sensor, wherein the sensor data acquisition parameters include sensor data sampling frequency and triggering mode;

[0077] Based on the layout strategy of the temperature sensor array, the deployment strategy of the vibration sensor and the sensor data acquisition parameters, a sensor network deployment strategy is formed and sent to the multi-source sensor network system, so that the multi-source sensor network system is deployed based on the sensor network deployment strategy.

[0078] In this embodiment, in order to ensure the quality and reliability of the detection data, it is necessary to first arrange the multi-channel ultrasonic array detection system and the multi-source sensor network system reasonably. Specifically, first obtain the ultrasonic detection requirements, which include key information such as the detection area, detection accuracy requirements, and detection frequency. At the same time, it is necessary to obtain environmental monitoring requirements, including information such as the temperature monitoring range and vibration monitoring frequency band, as well as obtain the structural characteristics of the rail transit equipment, such as the geometric dimensions of the equipment, material characteristics, etc.

[0079] When determining the structural strategy of a multi-channel ultrasonic array, it is necessary to comprehensively consider the coverage requirements and spatial constraints of the detection area. For example, for the detection of rail transit wheels, a linear array arrangement can be used, and the number of array elements can be selected as 16, 32 or 64. The array element spacing is usually set to between half a wavelength and one wavelength, such as less than 0.5mm, and the array element size can be λ / 2) to achieve better sound field coverage. Ultrasonic excitation parameters include center frequency (such as 2-5MHz), pulse width, emission sound pressure (such as 5MPa) and repetition frequency. The selection of these parameters needs to balance the requirements of detection depth and resolution to ensure the penetration ability and resolution of the signal.

[0080] In addition, the configuration of the ultrasonic signal acquisition parameters is also crucial. In an embodiment of the present application, the ultrasonic signal sampling frequency can be set to 5-10 times the center frequency to meet the requirements of the Nyquist sampling theorem. The quantization accuracy is generally selected to be 14 bits or 16 bits to ensure that the dynamic range of the collected signal meets the detection requirements. For example, when the ultrasonic center frequency is 5MHz, the sampling frequency can be set to 30MHz and the quantization accuracy can be set to 14 bits. The ultrasonic signal acquisition parameters also include data cache size, etc. Optionally, this embodiment establishes a real-time monitoring mechanism to ensure the reliability of data acquisition by setting a signal-to-noise ratio threshold (not less than 20dB) and a waveform quality index.

[0081] Furthermore, based on the structural strategy of the multi-channel ultrasonic array, the ultrasonic signal acquisition parameters and the ultrasonic excitation parameters, an ultrasonic array layout strategy is formed and sent to the multi-channel ultrasonic array detection system, so that the multi-channel ultrasonic array detection system is arranged based on the ultrasonic array layout strategy to achieve the collection of high-quality and high-precision ultrasonic reflection signals.

[0082] It should be noted that the layout strategy of the temperature sensor array needs to take into account the hot spot distribution characteristics of rail transit equipment. In this embodiment, the temperature sensor array can be arranged in key parts of the equipment, such as bearings, motors and other parts that are prone to heat. Exemplarily, the temperature sensor can be a sensor with a temperature measurement range of -40°C to 85°C and an accuracy better than ±0.5°C. For the deployment of vibration sensors, it is necessary to consider the vibration transmission path and modal characteristics of the equipment, and the sensors are usually arranged in positions where the vibration response is obvious. For example, for the monitoring of rail transit vehicle bearings, vibration sensors can be arranged in the radial and axial directions of the bearing seat. Exemplarily, the vibration sensor can be an acceleration sensor with a measurement range of ±50g and a frequency response of 0-5kHz.

[0083] In terms of the configuration of sensor data acquisition parameters, the sampling frequency of temperature data can be relatively low, such as 1Hz-10Hz, while the sampling frequency of vibration data needs to be higher, usually between a few kHz and tens of kHz. The trigger mode can be selected as a timed trigger or a conditional trigger according to actual needs. For example, when the device speed reaches a certain threshold, data acquisition is triggered. The sensor data acquisition parameters also include data packaging format, etc., to achieve synchronous acquisition of environmental data.

[0084] Optionally, this embodiment can realize the self-diagnosis function of the multi-source sensor network system, and ensure the stability and reliability of data collection through periodic testing and status monitoring. In addition, a data preprocessing module is constructed to perform preliminary filtering and calibration on the collected environmental data to provide a high-quality data basis for subsequent analysis.

[0085] Therefore, in order to ensure data quality, this application also needs to monitor the working status of each sensor in real time, diagnose and handle possible sensor failures in a timely manner, and ensure the reliability and continuity of data collection. After completing the above steps, a complete multi-source detection data stream is formed as the data basis for subsequent processing and analysis. Thus, the problem of a single detection method is solved, multi-source detection data is collected, and the comprehensiveness and accuracy of detection are improved.

[0086] S102: performing data preprocessing on the detection data to obtain preprocessed data, and extracting a feature vector based on the preprocessed data.

[0087] In this application, data cleaning, noise reduction and standardization are achieved by performing data preprocessing and feature extraction on the detection data.

[0088] S103: constructing a few-step diffusion model based on the feature vector and the preprocessed data, and performing data enhancement on the detection data through the few-step diffusion model to obtain synthetic data.

[0089] S104: construct a dense reward difference learning framework based on the synthetic data, optimize the dense reward difference learning framework, and obtain a detection model.

[0090] S105 . Construct a jump data diffusion model based on the detection model, and perform robustness analysis on the detection data based on the jump data diffusion model to obtain a robustness analysis result.

[0091] S106. Based on the robustness analysis result, a preset implicit neural representation network is used to extract the time continuity features in the detection data, and a time series anomaly detection is performed on the time series anomaly features to obtain a time series anomaly detection result.

[0092] S107, performing multi-model data fusion on the abnormal data detection result, the robustness analysis result and the time series abnormality detection result to obtain a detection warning result, wherein the abnormal data detection result is obtained after the detection model performs abnormality detection on the detection data.

[0093] The rail transit nondestructive testing method provided in this embodiment solves the problem of a single detection method by acquiring ultrasonic reflection signals from a multi-channel ultrasonic array detection system for rail transit equipment and temperature data and vibration data from a multi-source sensor network system for rail transit equipment, thereby realizing the collection of multi-source detection data and improving the comprehensiveness and accuracy of the detection. Subsequently, by preprocessing and extracting features of the multi-source detection data and using a few-step diffusion model for data enhancement, the fusion capability of the multi-source detection data is improved, and the training requirements of subsequent models can be met. Furthermore, by constructing a few-step diffusion model and a dense reward difference learning framework, the accuracy of abnormal data detection is improved. By constructing a jump data diffusion model, the robustness of the detection results of the detection model is improved, and the method can be applied to different working conditions and environmental conditions. By extracting time continuity features through an implicit neural representation network for time series anomaly detection, a deep analysis of the time series features of the detection data is realized, thereby performing multi-model data fusion on the abnormal data detection results, the robustness analysis results, and the time series anomaly detection results, and improving the detection accuracy and the reliability of the detection warning results.

[0094] In some embodiments, step S102 includes:

[0095] Performing wavelet transform processing on the ultrasonic reflection signal, processing the high-frequency coefficients obtained after the wavelet transform, and reconstructing the signal based on the processed high-frequency coefficients;

[0096] Performing median filtering on the temperature data to eliminate mutation points in the temperature data, and performing low-pass filtering on the temperature data to eliminate high-frequency noise in the temperature data;

[0097] Performing bandpass filtering on the vibration data and processing abnormal value points in the vibration data;

[0098] Based on the timestamp information, the processed ultrasonic reflection signal, the processed temperature data and the processed vibration data are aligned and standardized to obtain preprocessed data.

[0099] In this embodiment, the ultrasonic reflection signal is processed by wavelet transform method, and specifically Daubechies wavelet or Symlet wavelet can be selected. The signal can be decomposed into sub-signals of different frequency bands by wavelet transform, and the high-frequency coefficients are threshold processed to remove noise, and then the signal is reconstructed. For example, the high-frequency coefficients can be processed by soft threshold method, and the threshold can be determined based on the noise variance estimation. Exemplarily, the db4 wavelet basis function is selected, and 4-layer wavelet decomposition is performed. The high-frequency coefficients are processed by soft threshold method, and the signal-to-noise ratio is improved by not less than 10dB after the signal is reconstructed.

[0100] For temperature data, the window size of the median filter is usually selected to be 3-5 data points to effectively remove mutation points without over-smoothing the effective signal. Low-pass filtering can use a Butterworth filter, and the cutoff frequency is set according to the actual characteristics of temperature changes, usually between 0.1Hz-1Hz. For example, a 5-point median filter is used to eliminate mutation points, and a Butterworth low-pass filter (cutoff frequency 0.1Hz) is used to remove high-frequency noise to ensure the smoothness of temperature data.

[0101] For vibration data, the bandpass filter of the vibration data needs to be designed according to the characteristic frequency of the equipment. For example, for bearing vibration signals, the passband of the bandpass filter can be set within the range of 0.5-2 times the fault characteristic frequency. For the processing of outlier points, the 3σ criterion can be used for judgment and processing. Exemplarily, bandpass filtering (passband is 10Hz-2kHz) is implemented on the vibration data to eliminate low-frequency drift and high-frequency interference, and an adaptive threshold method is used to identify and process outlier points.

[0102] Furthermore, optionally, data alignment and synchronization are performed for the processed ultrasonic reflection signals, processed temperature data, and processed vibration data, and data from different sources are aligned to a unified time base based on timestamp information (with an accuracy better than 1ms). Subsequently, data standardization is performed, and the Z-score standardization method is used to make the mean of all data 0 and the standard deviation 1, which is convenient for subsequent analysis. Furthermore, a data quality assessment mechanism is established, and the quality of the preprocessing results is assessed by calculating indicators such as the signal-to-noise ratio and the effective data rate to ensure data availability.

[0103] Therefore, this embodiment performs data preprocessing through a multi-stage signal processing method to achieve noise reduction and standardization of the original detection data, thereby improving the quality of the detection data.

[0104] In some embodiments, step S102 further includes:

[0105] Performing time domain feature extraction and frequency domain feature extraction on the preprocessed ultrasonic reflection signal to obtain ultrasonic reflection signal feature data;

[0106] Extracting temperature characteristic data of the preprocessed temperature data, wherein the temperature characteristic data includes temperature gradient, temperature field distribution characteristics and temperature fluctuation characteristics;

[0107] Extracting vibration feature data of the preprocessed vibration data, wherein the vibration feature data includes time domain statistical features, frequency domain features, and time domain joint features;

[0108] The ultrasonic reflection signal characteristic data, the temperature characteristic data and the vibration characteristic data are screened and evaluated to obtain a characteristic vector.

[0109] In this embodiment, for the preprocessed ultrasonic reflection signal corresponding to the ultrasonic reflection signal, the time domain features include statistical features such as peak value, mean value, variance, skewness, kurtosis, etc., and also include zero crossing rate, envelope curve features, energy distribution features of the time domain waveform, etc. The frequency domain features include spectrum energy distribution, main frequency component, spectrum center of gravity, bandwidth, energy proportion of each frequency band (dividing the spectrum into 10 equally divided frequency bands), frequency domain entropy value, etc. Specifically, the spectrum can be obtained by fast Fourier transform, and then features such as frequency band energy ratio and spectrum center of gravity can be extracted.

[0110] For temperature characteristic data, the extraction of temperature characteristic data needs to consider the two dimensions of space and time. For example, the temperature gradient characteristics can be obtained by calculating the temperature difference between adjacent measuring points, the temperature field distribution characteristics can be extracted by the principal component analysis method, and the temperature fluctuation characteristics can be characterized by calculating the temperature change rate and fluctuation period.

[0111] For vibration characteristic data, time domain statistical features include root mean square value, peak factor, margin factor, etc., frequency domain features include frequency band energy, harmonic ratio, etc., and time-frequency joint features can be obtained through methods such as wavelet packet decomposition or Hilbert-Huang transform.

[0112] When screening and evaluating features, a method combining correlation analysis and importance evaluation can be used. First, calculate the correlation coefficient between features, remove highly correlated redundant features, and then evaluate the importance of features through methods such as random forests, and retain features that contribute more to the detection results. Exemplarily, a feature selection mechanism is constructed, based on the mutual information criterion and Fisher score, the most discriminative feature subset is selected from the initial feature set (the feature dimension is controlled at 20-30), the final feature vector is generated, and a feature evaluation system is established. By calculating the stability index and discrimination index of the feature, the quality of the feature extraction results is evaluated to ensure the validity and reliability of the feature.

[0113] In this embodiment, the feature extraction link is of great significance for improving the detection accuracy and reliability. First, feature extraction is performed on the preprocessed ultrasonic reflection signal. In terms of time domain feature extraction, it mainly includes the statistical characteristics of the signal, such as the peak amplitude can reflect the size of the defect, the root mean square value can characterize the energy level of the signal, the peak factor and margin factor can reflect the waveform characteristics of the signal, and the skewness and kurtosis can describe the degree of deviation of the signal amplitude distribution. In addition, it also includes the envelope characteristics of the signal. The instantaneous amplitude and phase information of the signal can be obtained through the Hilbert transform, which is important for identifying the reflection characteristics of ultrasonic waves at different medium interfaces. In terms of frequency domain feature extraction, the spectrum of the signal is first obtained through fast Fourier transform, and then the characteristics such as band energy ratio, main frequency component, and spectrum center of gravity are extracted. These characteristics can reflect the frequency composition and energy distribution characteristics of the signal.

[0114] For preprocessing temperature data, features need to be extracted from both spatial and temporal dimensions. Among them, the temperature gradient feature is obtained by calculating the temperature difference between adjacent measuring points, which can reflect the change trend of the temperature field and the heat conduction characteristics. The temperature field distribution feature uses the principal component analysis method to map the high-dimensional temperature field data to the low-dimensional feature space and extract the main temperature distribution pattern. The temperature fluctuation features include the temperature change rate, fluctuation period and fluctuation amplitude, etc. These features can reflect the dynamic temperature change law during the operation of the equipment.

[0115] For preprocessing vibration data, it is necessary to extract time domain statistical features, frequency domain features and time-frequency joint features. Among them, time domain statistical features include root mean square value, peak factor, margin factor, waveform factor, etc. These features can characterize the intensity and waveform characteristics of the vibration signal. Frequency domain features are obtained by spectral analysis of the vibration signal, including energy distribution of each frequency band, harmonic ratio, side band ratio, etc. These features are of great significance for identifying the fault type of the equipment. Time-frequency joint features are obtained through methods such as wavelet packet decomposition or Hilbert-Huang transform, which can describe the local characteristics of the vibration signal in the time and frequency dimensions.

[0116] After completing the above feature extraction, the feature data needs to be screened and evaluated to reduce the redundancy of the features and improve their contribution to the detection results. First, by calculating the Pearson correlation coefficient between the features, highly correlated feature pairs are identified, and one of the features with better physical interpretation is retained. Then, the importance of the features is evaluated by random forest and other methods, and the contribution of each feature to the model performance is calculated. Specifically, the features can be sorted by feature importance scores, and feature subsets with higher scores can be selected. In addition, domain knowledge can be combined to retain features that have clear physical meanings and are important indicators for fault diagnosis. Through this feature screening and evaluation mechanism, a feature vector with moderate dimensions and rich information is finally obtained, providing a reliable data basis for subsequent fault detection and diagnosis.

[0117] It should be noted that the process of feature extraction and screening needs to consider the characteristics of the actual application scenario. For example, for high-speed rail transit equipment, it may be necessary to pay more attention to the high-frequency characteristics of the vibration signal; for components that are prone to overheating failures, it is necessary to focus on extracting abnormal characteristics of the temperature field. Therefore, in practical applications, the feature extraction strategy can be appropriately adjusted according to specific needs to obtain the best detection effect. This embodiment extracts and selects features from multi-source detection data to achieve data dimensionality reduction and information extraction, which facilitates multi-source data fusion.

[0118] In some embodiments, constructing a few-step diffusion model based on the feature vector and the preprocessed data includes:

[0119] Based on the feature vector, construct a few-step diffusion initial model, wherein the few-step diffusion initial model includes a basic model architecture, a noise scheduling strategy, a sampling mechanism, a loss function, and an optimization strategy;

[0120] Dividing the feature vector and the preprocessed data into a training set, a validation set and a test set;

[0121] Using the training set to train the few-step diffusion initial model;

[0122] Using the validation set to evaluate and optimize the trained initial model of few-step diffusion;

[0123] The optimized few-step diffusion initial model is tested using the test set to obtain the few-step diffusion model.

[0124] In this embodiment, in order to solve the problem of insufficient model training data, the detection data is enhanced by constructing a few-step diffusion model to achieve the generation of high-quality samples and the optimization of data distribution. First, a few-step diffusion initial model is constructed based on the obtained feature vector. The basic model architecture of the few-step diffusion initial model uses U-Net as the backbone network, including two main parts: an encoder and a decoder. Among them, the encoder is composed of a plurality of downsampling blocks, each of which contains two convolutional layers, a batch normalization layer, and a ReLU activation function, and multi-scale features are extracted by gradually reducing the spatial resolution of the feature map. The decoder includes a corresponding upsampling block, which gradually restores the spatial resolution of the feature map through transposed convolution or upsampling operations. It should be noted that a jump connection is set between the encoder and the decoder, which can effectively retain the detail information and prevent the information from being lost during multiple downsampling processes.

[0125] For example, for the basic architecture of the few-step diffusion initial model, a backbone network based on U-Net is constructed, including a 4-layer encoder and a 4-layer decoder. Each encoder layer contains two 3×3 convolutional layers (the number of channels is 64, 128, 256, and 512, respectively) and a maximum pooling layer. Each decoder layer contains a transposed convolutional layer and two 3×3 convolutional layers. Skip connections are added between the layers to ensure the retention of detail information. In addition, the dimensions of the input layer and the output layer match the dimensions of the feature vector.

[0126] In addition, the noise scheduling strategy of the few-step diffusion initial model adopts cosine scheduling. Specifically, during the training process, the noise level gradually decreases from high to low, and the decreasing curve follows the cosine function. This scheduling method can improve the computational efficiency while ensuring the sampling quality. Exemplarily, the progressive noise addition mechanism sets the number of diffusion steps to 10 steps, and the noise scale of each step increases linearly in the range of β=0.1 to 0.9, realizing adaptive noise intensity adjustment, dynamically adjusting noise parameters according to data characteristics, and establishing noise impact evaluation indicators, including signal-to-noise ratio and data fidelity.

[0127] In this embodiment, the sampling mechanism of the few-step diffusion initial model adopts a prediction-correction sampler, which gradually transforms the noise data into meaningful samples through multi-step prediction and correction operations. Specifically, the sampling mechanism implements a sampling strategy based on a predictor-corrector, uses importance sampling technology to accelerate the convergence process, designs a batch parallel sampling mechanism, improves generation efficiency, and constructs a sampling quality evaluation index system.

[0128] In this embodiment, the loss function of the few-step diffusion initial model adopts a combination of simple loss and variational lower bound, where the simple loss is used to measure the reconstruction error and the variational lower bound is used to constrain the distribution of the latent space. Exemplarily, the mean square error loss and the knowledge distillation loss are combined, a regularization term is added to control the model complexity, and contrastive learning loss is introduced to enhance feature representation, and the weight coefficients are set to 0.5, 0.3 and 0.2 respectively.

[0129] In this embodiment, the optimization strategy of the initial model of the few-step diffusion adopts the Adam optimizer, and the learning rate adopts cosine annealing scheduling to achieve better convergence effect. Exemplarily, the Adam optimizer is used, the initial value of the learning rate is set to 0.001, the cosine annealing scheduling is used, the cycle is set to 10 epochs, the early stopping mechanism is constructed, the tolerance is set to 5 epochs, the gradient clipping is realized, and the threshold is set to 5.0.

[0130] Before model training, it is necessary to reasonably divide the feature vector and preprocessed data. In the embodiment of the present application, the data set is divided into a training set, a validation set and a test set in a ratio of 6:2:2. The training set is used for parameter learning of the model, the validation set is used for performance evaluation and hyperparameter tuning of the model, and the test set is used to evaluate the generalization ability of the model. It should be noted that the consistency of data distribution should be maintained when dividing the data, and stratified sampling can be used for division to ensure that the proportions of samples of different categories in each subset are similar.

[0131] In the model training stage, small batch stochastic gradient descent is used for optimization. Specifically, the training data is divided into multiple batches, and the batch size can be set to 32 or 64. The specific value needs to be determined according to the GPU memory capacity and training efficiency. During the training process, the model parameters are continuously updated by the back propagation algorithm until the preset number of training rounds is reached or the convergence condition is met. For example, it can be set that when the loss on the validation set changes by less than 0.1% in 5 consecutive rounds, the model is considered to have converged. Exemplarily, this embodiment performs progressive training, adopts a curriculum learning strategy, and gradually increases the noise level. The noise scale is set to 0.1 in the initial training stage, and increases by 0.1 every 5 epochs. At least 100 iterations are performed at each noise level to monitor the quality indicators of the generated samples, including the FID score and the IS score.

[0132] In the model evaluation and optimization stage, the main focus is on the performance indicators on the validation set, including reconstruction error, quality score of generated samples, etc. Based on the evaluation results, the hyperparameters of the model can be adjusted, such as changing the number of network layers, adjusting the learning rate, modifying the batch size, etc. In addition, an early stopping strategy can be adopted, that is, when the performance indicators on the validation set no longer improve, stop training in time to prevent the model from overfitting.

[0133] Finally, the optimized model is tested on the test set to evaluate the generalization performance of the model. The test phase focuses on the performance of the model on unseen data, including the quality and diversity of the generated samples and the consistency with the real data distribution. It should be noted that the test set can only be used once, and the model cannot be further adjusted based on the test results to ensure the objectivity of the evaluation results.

[0134] Furthermore, after completing model training and testing, the trained few-step diffusion model can be used to enhance the test data. Specifically, the test data is first input into the model, and synthetic data with similar characteristics to the original data is generated through a sampling process. These synthetic data not only maintain the statistical characteristics of the original data, but also contain certain random changes, which can effectively expand the number and diversity of training samples. It should be noted that the generated synthetic data should be quality evaluated to eliminate those samples that are obviously unreasonable or of poor quality to ensure the reliability of the enhanced data set.

[0135] Therefore, this embodiment achieves data enhancement of the detection data by constructing a few-step diffusion model, thereby achieving the generation of high-quality samples and the optimization of data distribution. Among them, by optimizing the training of the few-step diffusion initial model, the optimization of model performance and the improvement of generation quality are achieved.

[0136] In some embodiments, step S104 includes:

[0137] Based on the synthetic data, construct a local feature reward function and a global feature reward function, wherein the local feature reward function is used to indicate a reward for the similarity of local features, and the global feature reward function is used to indicate a reward for the consistency of global features;

[0138] Based on the local feature reward function and the global feature reward function, construct a dense reward difference learning framework;

[0139] Based on the dense reward difference learning framework, calculating sample weights, and performing difference optimization based on contrastive learning of the synthetic data;

[0140] Optimizing parameters of the dense reward difference learning framework based on the sample weights;

[0141] The dense reward difference learning framework with optimized parameters is evaluated and optimized to obtain the detection model.

[0142] Optionally, the local feature reward function Rlocal can be defined as:

[0143] ;

[0144] Among them, fi represents the feature representation of the i-th sample, N(i) represents the neighbor set of the i-th sample, fj represents the feature representation of the j-th sample, and sim() represents the cosine similarity. Among them, the samples in the local feature reward function are derived from synthetic data.

[0145] The global feature reward function Rglobal can be defined as:

[0146] ;

[0147] in, Represents the mean of all sample features.

[0148] In calculating sample weights When , it can be calculated based on the sparseness and representativeness of the sample:

[0149] ;

[0150] in, is a balancing factor, usually 0.5, sparsity(fi) represents the sparsity of the feature representation of the i-th sample, and representativeness(fi) represents the representativeness of the feature representation of the i-th sample.

[0151] In this embodiment, the construction of a dense reward difference learning framework is of great significance for improving the performance of the detection model. First, a local feature reward function and a global feature reward function are constructed. Among them, the local feature reward function is mainly used to measure the similarity of samples in the local feature space, and the reward value is obtained by calculating the feature similarity between each sample and its neighboring samples. This local feature similarity calculation can help the model better understand the local structural characteristics of the data. The global feature reward function focuses on the distribution of samples in the overall feature space, and obtains the reward value by calculating the difference between the sample features and the overall feature mean, which helps to maintain the diversity of sample features and prevent the model from falling into a local optimal solution.

[0152] When constructing a dense reward difference learning framework, it is necessary to organically combine the local feature reward function and the global feature reward function. Specifically, first design a basic deep neural network as a feature extractor to extract effective features from the input data. Then, calculate the local reward value and the global reward value based on the extracted features, and fuse the two rewards in a weighted manner to form the final reward signal. In addition, it is also necessary to design a corresponding loss function to guide the training process of the model so that it can simultaneously focus on the similarity of local features and the consistency of global features.

[0153] In terms of sample weight calculation and difference optimization, we first need to evaluate the importance of each sample. Specifically, the weight value can be calculated based on the sparsity and representativeness of the sample. Sparsity reflects the uniqueness of the sample in the feature space, and representativeness reflects the contribution of the sample to the overall data distribution. After obtaining the sample weight, difference optimization is performed, that is, through comparative learning, the model can learn the subtle differences between samples. In this process, samples with high weights will receive more attention, thereby improving the model's ability to identify key samples.

[0154] Based on the obtained sample weights, the dense reward difference learning framework is optimized. In this stage, the model parameters are updated in a weighted manner so that the model pays more attention to samples with higher weights during training. Specifically, targeted parameter optimization can be achieved by adjusting the contribution ratio of different samples in the loss function or weighting the gradient during back propagation. It should be noted that the parameter optimization process should adopt an appropriate learning rate scheduling strategy to ensure that the model can converge stably.

[0155] Finally, the optimized framework is systematically evaluated and optimized. The evaluation process focuses on the performance indicators of the model on the validation set, including detection accuracy, recall rate, precision, etc. Based on the evaluation results, the various components of the framework can be further adjusted, such as modifying the network structure, adjusting the weight of the reward function, optimizing the calculation method of sample weights, etc. In this process, multiple iterations are required until the model performance reaches the expected goal. It should be noted that in order to ensure the robustness of the model, the evaluation data should include various possible working conditions and fault types to ensure that the obtained detection model has good generalization ability.

[0156] In addition, after obtaining the final detection model, model compression and acceleration optimization are required to meet the needs of actual deployment. For example, through technologies such as knowledge distillation, model pruning or quantization, the computational complexity and storage overhead of the model can be reduced while ensuring detection performance, so that it can run efficiently in a resource-constrained environment.

[0157] Therefore, this embodiment constructs a dense reward difference learning framework by constructing a local feature reward function and a global feature reward function, thereby constructing a detection model, and realizing accurate identification and evaluation of abnormal samples in the detection data using the detection model, thereby improving the accuracy of abnormal identification of the detection data.

[0158] In some embodiments, step S105 includes:

[0159] Based on the detection model, a multi-scale feature extraction network is constructed, wherein the multi-scale feature extraction network includes a skip connection structure in which a plurality of feature scale layers arranged in cascade are arranged between adjacent feature scale layers;

[0160] Constructing a feature fusion mechanism, wherein the feature fusion mechanism is used to indicate that a preset channel attention mechanism and a spatial attention mechanism are used to perform feature fusion on the detection data;

[0161] Constructing an anti-interference network, wherein the anti-interference network is used to instruct to perform feature alignment on the detection data using a preset spatial transformation network;

[0162] Based on the multi-scale feature extraction network, the feature fusion mechanism and the anti-interference network, a jump data diffusion model is constructed;

[0163] Performing system performance evaluation and optimization on the jump data diffusion model;

[0164] Based on the optimized jump data diffusion model, a robustness analysis is performed on the detection data to obtain a robustness analysis result.

[0165] It should be noted that the multi-scale feature extraction network of the jump data diffusion model adopts a pyramid structure, which contains 3-5 feature scale layers, and the size of the feature map of each layer is halved. The jump connection between adjacent scale layers adopts the form of residual connection, which can effectively avoid the gradient disappearance problem.

[0166] The channel attention in the feature fusion mechanism adopts the Squeeze-and-Excitation structure, and the spatial attention adopts the self-attention mechanism. Specifically, the channel attention module first performs global average pooling on the feature map, and then learns the channel weights through two fully connected layers. The spatial attention module obtains the spatial weights by calculating the correlation between different positions of the feature map.

[0167] The anti-interference network adopts the idea of ​​spatial transformation network, which includes three main components: localization network, grid generator and sampler. The localization network predicts the spatial transformation parameters, the grid generator generates the sampling grid based on the transformation parameters, and the sampler resamples the feature map according to the sampling grid. This mechanism can adaptively align and correct features and improve the robustness of the model to posture changes.

[0168] In this embodiment, the construction of the jump data diffusion model is a key link in realizing robustness detection. First, a multi-scale feature extraction network is constructed, which adopts a pyramid structure and includes 3 to 5 feature scale layers. Each feature scale layer gradually reduces the spatial resolution of the feature map through convolution operations and downsampling operations, while increasing the number of feature channels to obtain feature representations of different scales. A jump connection is set between adjacent feature scale layers in the form of a residual connection. This design can effectively alleviate the gradient vanishing problem and can retain more detail information during the feature extraction process. It should be noted that each feature scale layer contains multiple residual blocks, each of which is composed of a convolution layer, a batch normalization layer and an activation function. This deep residual structure can improve the feature extraction capability of the network.

[0169] In terms of the construction of the feature fusion mechanism, it mainly includes two key components: channel attention and spatial attention. The channel attention mechanism adopts the Squeeze-and-Excitation structure. First, the statistical features of each channel are obtained through global average pooling, and then the relationship between channels is learned through a two-layer fully connected network, and finally the channel weight is generated. This mechanism can help the model identify and strengthen important feature channels. The spatial attention mechanism adopts a self-attention structure to generate spatial weights by calculating the correlation between different positions of the feature map. In this way, the model can adaptively focus on important spatial areas and improve the pertinence of feature extraction. In addition, a multi-level fusion strategy is also adopted in the feature fusion process, that is, feature fusion is performed between different feature scale layers to make full use of multi-scale feature information.

[0170] The construction of the anti-interference network adopts the idea of ​​spatial transformation network. First, the localization network is used to predict the spatial transformation parameters, which describe the geometric transformations that the feature map needs to undergo, such as translation, rotation, and scaling. Then, the grid generator generates a sampling grid based on the predicted transformation parameters, which defines the sampling positions on the feature map. Finally, the sampler resamples the feature map according to the sampling grid to obtain a feature map that has undergone spatial transformation. Through this mechanism, the model can adaptively align and correct features, effectively reducing the interference caused by posture changes. It should be noted that the learning process of the spatial transformation parameters is end-to-end and does not require additional supervision information.

[0171] Based on the above three core components, a complete jump data diffusion model is constructed. The model adopts an encoder-decoder structure, in which the encoder contains a multi-scale feature extraction network to obtain multi-scale feature representations. The decoder reconstructs the representation of the input data through step-by-step upsampling and feature fusion. Multiple layers of jump connections are set between the encoder and the decoder, which can help the decoder better utilize the feature information extracted by the encoder. The training process of the model adopts an end-to-end approach, and the loss function includes reconstruction loss and regularization terms. The reconstruction loss is used to ensure the accuracy of the reconstruction result, and the regularization term is used to constrain the complexity of the model.

[0172] In the system performance evaluation and optimization stage, the performance of the model under various working conditions, including different noise levels, different interference types, and different failure modes, is first evaluated. Evaluation indicators include reconstruction error, feature extraction capability, anti-interference capability, etc. Based on the evaluation results, the various components of the model can be optimized in a targeted manner, such as adjusting the network structure, modifying the attention mechanism, optimizing the loss function, etc. In addition, the computational efficiency and resource consumption of the model need to be considered. If necessary, optimization can be performed using technologies such as model compression or knowledge distillation.

[0173] Finally, the optimized model is used to perform robustness analysis on the detection data. Specifically, different types of interference data are first constructed, such as adding noise, performing geometric transformations, changing lighting conditions, etc. Then, the performance of the model on these interference data is evaluated, including the stability of feature extraction, the consistency of detection results, etc. Through this systematic robustness analysis, the anti-interference ability of the model can be comprehensively evaluated and possible performance bottlenecks can be identified. It should be noted that the results of the robustness analysis can serve as an important basis for further optimizing the model and help improve the reliability of the model in practical applications.

[0174] Therefore, this embodiment constructs a jump data diffusion model through a multi-scale feature extraction network, a feature fusion mechanism and an anti-interference network, which can be used to perform data enhancement and feature extraction on the detection data, thereby meeting the robustness analysis requirements for the detection data and achieving the robustness of the analysis of the detection data. It solves the problem that the existing detection models are insufficiently robust and difficult to adapt to different working conditions and environmental conditions, achieves enhanced adaptability and stability of data under different working conditions, and improves the comprehensiveness and accuracy of the analysis of the detection data.

[0175] In some embodiments, step S106 includes:

[0176] Based on the robustness analysis result, constructing an implicit neural representation network, wherein the implicit neural representation network includes a time continuity encoder, a temporal dependency model and an anomaly detection discriminator;

[0177] Extracting a temporal continuity feature in the detection data based on the implicit neural representation network, and calculating an abnormality score of the temporal continuity feature;

[0178] The anomaly score of the time continuity feature is judged based on a preset anomaly threshold to obtain a corresponding time series anomaly detection result.

[0179] In this embodiment, the temporal continuity encoder of the implicit neural representation network adopts a multi-layer perceptron structure, the number of neurons in the input layer is the same as the feature dimension, and the hidden layer adopts the ReLU activation function. The temporal dependency model adopts a long short-term memory network (LSTM) structure based on the attention mechanism, which can capture long-term and short-term temporal dependencies. The anomaly detection discriminator adopts a reconstruction error-based method, first reconstructing the input features, and then calculating the reconstruction error as the anomaly score.

[0180] When determining the abnormality threshold, it can be set based on the abnormality score distribution of normal samples. Then, when it is detected that the abnormality score exceeds the abnormality threshold, it is considered that there is a time series abnormality, and the corresponding time series abnormality detection result is output.

[0181] In this embodiment, time series anomaly detection is implemented based on an implicit neural representation network. First, an implicit neural representation network is constructed, which includes three main components: a time continuity encoder, a time series dependency model, and an anomaly detection discriminator. Among them, the time continuity encoder adopts a multi-layer perceptron structure, and maps the input features to an implicit feature space through multiple fully connected layers and nonlinear activation functions. It should be noted that the design of the encoder needs to consider the temporal continuity of the features, so a time embedding layer is added to the network structure to capture the temporal information of the features. The temporal dependency model adopts a long short-term memory network structure based on an attention mechanism, which can effectively model long-term and short-term temporal dependencies. Specifically, through the gating mechanism and memory units, the model can selectively memorize and forget historical information, thereby capturing important patterns in the data sequence. The anomaly detection discriminator adopts a reconstruction-based approach to achieve anomaly detection by reconstructing the input features and calculating the reconstruction error.

[0182] In the stage of temporal continuity feature extraction and anomaly score calculation, the detection data is first organized into time series and sent to the implicit neural representation network for processing. The temporal continuity encoder converts the input features into implicit representations, and the temporal continuity information of the features is retained in this process. Then, the temporal dependency model models the implicit representations and captures the temporal dependencies in the feature sequence. By comparing the differences between the features at the current moment and the features predicted by the model, potential abnormal patterns can be identified. In addition, the multi-scale characteristics of the features need to be considered, so a multi-scale analysis method is used in the feature extraction process, that is, features are extracted at different time scales and anomaly scores are calculated. It should be noted that the calculation of the anomaly score should not only consider the reconstruction error, but also the temporal continuity and prediction deviation of the features, so that the degree of abnormality of the data can be more comprehensively evaluated.

[0183] Finally, the anomaly score is judged based on the preset anomaly threshold to obtain the time series anomaly detection result. The setting of the threshold is a key link, and a balance needs to be struck between the false alarm rate and the missed alarm rate. In an embodiment of the present application, a method based on the statistical characteristics of normal samples is used to determine the threshold. Specifically, a reasonable threshold can be set by analyzing the distribution characteristics of the anomaly score under normal operating conditions. In addition, considering that the characteristics of different types of anomalies may be different, different thresholds can be set for different feature dimensions. This adaptive threshold setting method can improve the accuracy of detection. It should be noted that in actual applications, the setting of the threshold also needs to consider the specific application scenarios and safety requirements, and can be dynamically adjusted according to the actual operating conditions.

[0184] In addition, in order to improve the reliability of the detection results, the embodiment of the present application also introduces a time continuity judgment mechanism. Specifically, the abnormal detection result is triggered only when the abnormal state lasts for a certain period of time, which can effectively filter out false alarms caused by instantaneous fluctuations. At the same time, different duration requirements are set for abnormalities of different severity. This hierarchical judgment mechanism can better guide actual maintenance decisions. It should be noted that the judgment threshold of time continuity can also be adjusted according to actual operating experience to adapt to different application requirements.

[0185] Finally, the system will also record and analyze the historical results of anomaly detection, and continuously optimize the detection strategy and parameter settings by statistically analyzing the laws and characteristics of anomaly occurrence. For example, the temporal distribution and spatial distribution of anomalies and their correlation with the operating status of the equipment can be analyzed. These analysis results can provide important references for predictive maintenance. In addition, the anomaly detection model can be regularly evaluated and updated based on historical data to ensure the continued effectiveness of detection performance.

[0186] Therefore, this embodiment extracts the time continuity features in the detection data through the implicit neural representation network, so as to analyze whether there are timing anomalies in the detection data, realize in-depth analysis of the timing features of the detection data, and improve the accuracy of the detection analysis.

[0187] In some embodiments, step S107 includes:

[0188] Assigning corresponding weight coefficients to the abnormal data detection result, the robustness analysis result and the timing anomaly detection result;

[0189] Performing weighted fusion on the abnormal data detection result, the robustness analysis result and the time series abnormality detection result according to the weight coefficient;

[0190] Output the corresponding detection and warning results based on the weighted fusion results.

[0191] In this embodiment, the weight coefficients corresponding to the abnormal data detection results, the robustness analysis results, and the timing anomaly detection results are determined respectively, and then weighted calculations are performed to output the corresponding detection and early warning results. It should be noted that the output of the detection and early warning results needs to consider the needs of the actual application scenarios. In the embodiment of the present application, the early warning results can be divided into multiple levels, such as normal, slight abnormality, severe abnormality, etc. Each level corresponds to a different processing strategy, such as continued monitoring, increasing the detection frequency, immediate shutdown inspection, etc.

[0192] In addition, in order to improve the reliability of the warning results, a time continuity judgment mechanism can also be introduced. For example, only when the abnormal state lasts for a certain period of time (such as 3 consecutive detection cycles) will the corresponding level of warning be triggered, which can effectively avoid false alarms caused by instantaneous fluctuations.

[0193] Therefore, this embodiment performs multi-model data fusion on the abnormal data detection results, robustness analysis results and time series anomaly detection results, thereby improving the detection accuracy and the reliability of the detection warning results.

[0194] An embodiment of the present application provides a rail transit non-destructive testing device, including a processor; a memory for storing a computer program; wherein, when the processor executes the computer program, the steps of the rail transit non-destructive testing method described in the above embodiment are implemented.

[0195] In an optional embodiment, a rail transit nondestructive testing device is provided, such as Figure 2 As shown, Figure 2The rail transit nondestructive testing device 200 shown includes: a processor 201 and a memory 203. The processor 201 and the memory 203 are connected, such as through a bus 202. Optionally, the rail transit nondestructive testing device 200 may also include a transceiver 204, which can be used for data interaction between the rail transit nondestructive testing device and other devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 204 is not limited to one, and the structure of the rail transit nondestructive testing device 200 does not constitute a limitation on the embodiments of the present application.

[0196] The processor 201 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. The processor 201 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0197] The bus 202 may include a path to transmit information between the above components. The bus 202 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 202 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0198] The memory 203 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, without limitation herein.

[0199] The memory 203 is used to store the computer program for executing the embodiment of the present application, and the execution is controlled by the processor 201. The processor 201 is used to execute the computer program stored in the memory 203 to implement the steps shown in the above method embodiment.

[0200] The above is a detailed description of the specific implementation method of the embodiment of the present application. Through the implementation of the above technical solution, efficient, accurate and reliable non-destructive testing of rail transit equipment can be achieved, providing a strong guarantee for the safe operation of rail transit.

Claims

1. A rail transit nondestructive testing method, characterized in that: include: Acquire detection data for rail transit equipment, the detection data including ultrasonic reflection signals from a multi-channel ultrasonic array detection system, and temperature data and vibration data from a multi-source sensor network system; Performing data preprocessing on the detection data to obtain preprocessed data, and extracting a feature vector based on the preprocessed data; Based on the feature vector and the preprocessed data, a few-step diffusion model is constructed, and data enhancement is performed on the detection data through the few-step diffusion model to obtain synthetic data; Based on the synthetic data, a dense reward difference learning framework is constructed, and the dense reward difference learning framework is optimized to obtain a detection model; Based on the detection model, a jump data diffusion model is constructed, and based on the jump data diffusion model, a robustness analysis is performed on the detection data to obtain a robustness analysis result; Based on the robustness analysis result, a preset implicit neural representation network is used to extract the time continuity feature in the detection data, and a time series anomaly detection is performed on the time series anomaly feature to obtain a time series anomaly detection result; Multi-model data fusion is performed on the abnormal data detection result, the robustness analysis result and the time series anomaly detection result to obtain a detection warning result, wherein the abnormal data detection result is obtained after the detection model performs anomaly detection on the detection data.

2. The method according to claim 1, characterized in that The performing data preprocessing on the detection data to obtain preprocessed data includes: Performing wavelet transform processing on the ultrasonic reflection signal, processing the high-frequency coefficients obtained after the wavelet transform, and reconstructing the signal based on the processed high-frequency coefficients; Performing median filtering on the temperature data to eliminate mutation points in the temperature data, and performing low-pass filtering on the temperature data to eliminate high-frequency noise in the temperature data; Performing bandpass filtering on the vibration data and processing abnormal value points in the vibration data; Based on the timestamp information, the processed ultrasonic reflection signal, the processed temperature data and the processed vibration data are aligned and standardized to obtain preprocessed data.

3. The method according to claim 2, characterized in that The preprocessed data includes preprocessed ultrasonic reflection signals, preprocessed temperature data, and preprocessed vibration data. The extracting of feature vectors based on the preprocessed data includes: Performing time domain feature extraction and frequency domain feature extraction on the preprocessed ultrasonic reflection signal to obtain ultrasonic reflection signal feature data; Extracting temperature characteristic data of the preprocessed temperature data, wherein the temperature characteristic data includes temperature gradient, temperature field distribution characteristics and temperature fluctuation characteristics; Extracting vibration feature data of the preprocessed vibration data, wherein the vibration feature data includes time domain statistical features, frequency domain features, and time-frequency joint features; The ultrasonic reflection signal characteristic data, the temperature characteristic data and the vibration characteristic data are screened and evaluated to obtain a characteristic vector.

4. The method according to claim 1, characterized in that The step of constructing a dense reward difference learning framework based on the synthetic data, optimizing the dense reward difference learning framework, and obtaining a detection model includes: Based on the synthetic data, construct a local feature reward function and a global feature reward function, wherein the local feature reward function is used to indicate a reward for the similarity of local features, and the global feature reward function is used to indicate a reward for the consistency of global features; Based on the local feature reward function and the global feature reward function, construct a dense reward difference learning framework; Based on the dense reward difference learning framework, calculating sample weights, and performing difference optimization based on contrastive learning of the synthetic data; Optimizing parameters of the dense reward difference learning framework based on the sample weights; The dense reward difference learning framework with optimized parameters is evaluated and optimized to obtain the detection model.

5. The method according to claim 1, characterized in that: The step of constructing a jump data diffusion model based on the detection model, and performing robustness analysis on the detection data based on the jump data diffusion model to obtain a robustness analysis result includes: Based on the detection model, a multi-scale feature extraction network is constructed, wherein the multi-scale feature extraction network includes a skip connection structure in which a plurality of feature scale layers arranged in cascade are arranged between adjacent feature scale layers; Constructing a feature fusion mechanism, wherein the feature fusion mechanism is used to indicate that a preset channel attention mechanism and a spatial attention mechanism are used to perform feature fusion on the detection data; Constructing an anti-interference network, wherein the anti-interference network is used to instruct to use a preset spatial transformation network to perform feature alignment on the detection data; Based on the multi-scale feature extraction network, the feature fusion mechanism and the anti-interference network, a jump data diffusion model is constructed; Performing system performance evaluation and optimization on the jump data diffusion model; Based on the optimized jump data diffusion model, a robustness analysis is performed on the detection data to obtain a robustness analysis result.

6. The method according to claim 1, characterized in that Based on the robustness analysis result, the preset implicit neural representation network is used to extract the time continuity feature in the detection data, and the time continuity feature is subjected to time series anomaly detection to obtain the time series anomaly detection result, including: Based on the robustness analysis result, constructing an implicit neural representation network, wherein the implicit neural representation network includes a time continuity encoder, a temporal dependency model and an anomaly detection discriminator; Extracting a temporal continuity feature in the detection data based on the implicit neural representation network, and calculating an abnormality score of the temporal continuity feature; The anomaly score of the time continuity feature is judged based on a preset anomaly threshold to obtain a corresponding time series anomaly detection result.

7. The method according to claim 1, characterized in that The performing multi-model data fusion on the abnormal data detection result, the robustness analysis result and the time series abnormality detection result to obtain the detection warning result includes: Assigning corresponding weight coefficients to the abnormal data detection result, the robustness analysis result and the timing anomaly detection result; Performing weighted fusion on the abnormal data detection result, the robustness analysis result and the time series abnormality detection result according to the weight coefficient; Output the corresponding detection and warning results based on the weighted fusion results.

8. The method according to claim 1, characterized in that: Before obtaining the detection data for the rail transit equipment, the method further includes: Obtaining ultrasonic testing requirements, environmental monitoring requirements and structural characteristics of the rail transit equipment; Based on the ultrasonic detection requirements and the structural characteristics of the rail transit equipment, determine the structural strategy and ultrasonic excitation parameters of the multi-channel ultrasonic array, wherein the structural strategy includes the number, spacing and arrangement of array elements; Configuring ultrasonic signal acquisition parameters of the multi-channel ultrasonic array, wherein the ultrasonic signal acquisition parameters include ultrasonic signal sampling frequency and quantization accuracy; Based on the structural strategy of the multi-channel ultrasonic array, the ultrasonic signal acquisition parameters and the ultrasonic excitation parameters, an ultrasonic array arrangement strategy is formed and sent to the multi-channel ultrasonic array detection system, so that the multi-channel ultrasonic array detection system is arranged based on the ultrasonic array arrangement strategy; Based on the environmental monitoring requirements and the structural characteristics of the rail transit equipment, determine the layout strategy of the temperature sensor array and the deployment strategy of the vibration sensor; Configuring sensor data acquisition parameters of the temperature sensor array and the vibration sensor, wherein the sensor data acquisition parameters include sensor data sampling frequency and triggering mode; Based on the layout strategy of the temperature sensor array, the deployment strategy of the vibration sensor and the sensor data acquisition parameters, a sensor network deployment strategy is formed and sent to the multi-source sensor network system, so that the multi-source sensor network system is deployed based on the sensor network deployment strategy.

9. The method according to claim 1, characterized in that: The constructing of a few-step diffusion model based on the feature vector and the preprocessed data includes: Based on the feature vector, construct a few-step diffusion initial model, wherein the few-step diffusion initial model includes a basic model architecture, a noise scheduling strategy, a sampling mechanism, a loss function, and an optimization strategy; Dividing the feature vector and the preprocessed data into a training set, a validation set and a test set; Using the training set to train the few-step diffusion initial model; Using the validation set to evaluate and optimize the trained initial model of few-step diffusion; The optimized few-step diffusion initial model is tested using the test set to obtain the few-step diffusion model.

10. A rail transit nondestructive testing device, characterized in that: include: processor; memory for storing computer programs; Wherein, when the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images