Fault detection method and device, equipment and storage medium

Abstract

The invention relates to the technical field of weighing sensor detection, and discloses a fault detection method, device and equipment and a storage medium, and the method comprises the steps: carrying out the independent analysis of multi-source data from different sensors, obtaining the analyzed data of each dimension, carrying out the correlation fusion of the analyzed data, generating a multi-dimensional fault feature matrix, and carrying out the calculation of the multi-dimensional fault feature matrix. And based on the multi-dimensional fault feature matrix, fault positioning is carried out on the weighing sensor through a deep learning model, and fault detection information is determined. According to the method, independent analysis, relevance fusion and deep learning model fault positioning are performed on the multi-source data, so that accurate detection and positioning of the fault of the weighing sensor are realized, the problem that multi-source data fusion and anomaly recognition under complex working conditions are difficult to deal with in the prior art is solved, and the comprehensiveness and accuracy of fault detection are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of weighing sensor detection technology, and in particular to a fault detection method, device, equipment and storage medium. Background Art

[0002] In modern industrial equipment, load cells serve as critical data acquisition units, and their operational status directly impacts equipment reliability and safety. However, with increasing equipment complexity and diverse operating conditions, load cell fault detection faces increasing challenges. Existing technologies often rely on the analysis of a single data source, which presents significant limitations in complex operating conditions and struggles with multi-source data fusion and anomaly identification. Summary of the Invention

[0003] The main purpose of this application is to provide a fault detection method, device, equipment and storage medium, aiming to solve the technical problem that existing weighing sensor fault detection technology relies on a single data source and is difficult to cope with multi-source data fusion and anomaly identification under complex working conditions.

[0004] To achieve the above objectives, the present application proposes a fault detection method, which includes: Independently analyze multi-source data from different sensors to obtain analyzed data in various dimensions; Performing correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix; Based on the multi-dimensional fault feature matrix, the fault location of the weighing sensor is performed through a deep learning model to determine the fault detection information.

[0005] In one embodiment, the multi-source data includes temperature data, humidity data, tilt angle data, housing integrity detection data, inert gas concentration data, interference information, and abnormal data fluctuation information. The step of independently analyzing the multi-source data from different sensors to obtain analyzed data in each dimension includes: Based on thermodynamic models and Fourier transform, the correlation analysis results between temperature anomalies and equipment operating status are obtained by extracting the periodic fluctuation characteristics of temperature data; By identifying the long-term change trend of humidity data, we can evaluate the impact of humidity data on the performance of load cells. The Kalman filter method is used to analyze the dynamic offset characteristics of the tilt angle data to obtain the potential damage analysis results of the abnormal tilt angle to the equipment structure; monitoring helium concentration data in the sealed environment in real time, and generating a shell integrity test result based on the helium concentration data; Utilizing a gas leakage model, by analyzing the gradient change of the inert gas concentration data, the gas leakage fault location result is obtained, wherein the gas leakage model is generated based on the diffusion law and concentration gradient change of the inert gas inside the equipment; By using adaptive filtering algorithm to separate the noise signal from the interference information, the effective signal data after the interference is eliminated is obtained; Based on the time series decomposition algorithm, by identifying the mutation points in the abnormal fluctuation information of the data, the correlation analysis results between abnormal fluctuation and fault are obtained; The analyzed data of each dimension is obtained by combining the correlation analysis results, the impact assessment results, the potential hazard analysis results, the shell integrity detection results, the positioning results, the valid signal data and the correlation analysis results.

[0006] In one embodiment, the step of monitoring helium concentration data in the sealed environment in real time and generating a housing integrity test result based on the helium concentration data includes: Real-time monitoring of helium concentration data in a sealed environment; Use pressure sensors to collect pressure fluctuation data in sealed environments, and combine them with pressure fluctuation analysis algorithms to extract abnormal pressure fluctuation characteristics of the pressure fluctuation data; Correlation analysis is performed on the helium concentration data and the pressure fluctuation data to identify a coordinated change pattern of abnormal increase in helium concentration and sudden change in pressure fluctuation; According to the coordinated change pattern, a shell integrity detection result is generated by determining whether a shell damage event occurs.

[0007] In one embodiment, the step of performing correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix includes: Based on a data alignment algorithm, performing time synchronization and spatial alignment on the analyzed data to obtain aligned data; A multi-dimensional fault feature matrix is constructed by assigning weights to the aligned data.

[0008] In one embodiment, the step of performing time synchronization and spatial alignment on the analyzed data based on a data alignment algorithm to obtain aligned data includes: Based on a data alignment algorithm, time-aligning the analyzed data to obtain preliminary time-aligned data; Eliminating the time offset of the preliminary time alignment data by a dynamic time warping algorithm to obtain time synchronization data; performing spatial alignment on the time synchronization data to obtain preliminary spatial alignment data; Based on the topological relationship of the device structure, the preliminary spatial alignment data is mapped to a unified fault coordinate system to obtain spatial alignment data; The time synchronization data and the spatial alignment data are combined to obtain aligned data.

[0009] In one embodiment, the step of locating the fault of the weighing sensor using a deep learning model based on the multi-dimensional fault feature matrix and determining the fault detection information includes: Inputting the multidimensional fault feature matrix into a pre-trained deep learning model to obtain spatial features and temporal features of the fault features; generating a comprehensive fault feature based on the spatial feature and the temporal feature; Based on the comprehensive fault characteristics, fault detection information of the weighing sensor is output according to a fully connected neural network.

[0010] In one embodiment, the step of outputting fault detection information of the weighing sensor according to a fully connected neural network based on the comprehensive fault characteristics includes: Inputting the comprehensive fault features into a fully connected neural network, and extracting high-order features of the fault through multi-layer nonlinear transformation; Based on the high-order features, the attention mechanism is used to dynamically assign the contribution weights of different features to fault localization to obtain a weight distribution result; According to the weight distribution result, the fault type is output by the fault classifier, and the location information in the high-order features is combined to generate a heat map of the fault location; Based on the fault type and the thermal map, fault detection information of the weighing sensor is output, wherein the fault detection information includes the fault type, the fault location, and the fault severity.

[0011] In addition, to achieve the above objectives, the present application also proposes a fault detection device, which includes: The data processing module is used to independently analyze multi-source data from different sensors to obtain analyzed data in various dimensions; A data fusion module is used to perform correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix; The information acquisition module is used to locate the fault of the weighing sensor based on the multi-dimensional fault feature matrix through a deep learning model to determine the fault detection information.

[0012] In addition, to achieve the above-mentioned purpose, the present application also proposes a fault detection device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the fault detection method described above.

[0013] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, the steps of the fault detection method described above are implemented.

[0014] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of the fault detection method described above are implemented.

[0015] The technical solution proposed in this application independently analyzes multi-source data from different sensors to obtain analyzed data in each dimension. This data is then correlated and fused to generate a multi-dimensional fault feature matrix. Based on this multi-dimensional fault feature matrix, a deep learning model is used to locate faults in the load cell and determine fault detection information. By independently analyzing multi-source data, correlatively fusing it, and locating faults using a deep learning model, this application achieves accurate detection and location of load cell faults. This solves the problem that existing technologies have difficulty coping with multi-source data fusion and anomaly identification under complex working conditions, significantly improving the comprehensiveness and accuracy of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 A flowchart of the first embodiment of the fault detection method of the present application is provided; Figure 2 A flowchart of the second embodiment of the fault detection method of this application is provided; Figure 3 A flowchart of the third embodiment of the fault detection method of this application is provided; Figure 4 This is a schematic diagram of the module structure of the fault detection device according to an embodiment of the present application; Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the fault detection method in the embodiment of the present application.

[0019] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0020] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0021] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0022] In the existing technology, weighing sensor fault detection mostly relies on the analysis of a single data source, which has obvious limitations under complex working conditions and is difficult to cope with multi-source data fusion and anomaly identification.

[0023] Therefore, in order to overcome the above-mentioned defects, the present application provides a solution, which realizes the accurate detection and location of load cell faults by independent analysis, correlation fusion and deep learning model fault location of multi-source data, solves the problem that the existing technology is difficult to cope with multi-source data fusion and anomaly identification under complex working conditions, and significantly improves the comprehensiveness and accuracy of fault detection.

[0024] It should be noted that the execution entity of each embodiment of the present application may be a computing service system with data processing, network communication, and program execution functions, such as an electronic system or fault detection system capable of implementing the above functions. The following embodiments are described using a fault detection system as an example (hereinafter referred to as the "system").

[0025] Based on this, the present application embodiment provides a fault detection method, referring to Figure 1 , Figure 1 This is a flowchart of the first embodiment of the fault detection method of the present application.

[0026] In this embodiment, the fault detection method includes steps S10 to S30: Step S10 , independently analyzing multi-source data from different sensors to obtain analyzed data in each dimension.

[0027] In complex industrial equipment, load cell fault detection is critical for ensuring safe operation. Traditional methods rely on the analysis of a single data source, making it difficult to fully reflect the equipment status. This is especially true when multi-source data (i.e., various types of data collected from different sensors, such as temperature, humidity, and tilt angle) coexist. This limits the accuracy and efficiency of fault detection. To address this issue, this proposal proposes a fault detection method based on multi-source data fusion and deep learning. By independently analyzing and correlating multi-source data, and using a deep learning model to locate faults, this method achieves accurate detection of load cell faults.

[0028] It's important to note that multi-source data reflects different aspects of equipment operation and therefore requires independent analysis. The goal of independent analysis is to extract the characteristics of each data type, such as periodic fluctuations in temperature data or long-term trends in humidity data. Through targeted analysis of each data type, we can obtain analyzed data for each dimension (i.e., characteristic data obtained through independent analysis of multi-source data, such as temperature anomalies and humidity trends).

[0029] In addition, it should be noted that a self-calibration mechanism can be introduced to dynamically adjust the weight of the weighing sensor by drawing on the bee swarm algorithm, imitating the collaborative calibration process of the biological group to improve data accuracy.

[0030] Step S20: performing correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix.

[0031] Understandably, while the analyzed data contains the characteristics of each data type, they may be inconsistent in time and space. The goal of correlation fusion is to integrate this data to generate a unified multi-dimensional fault feature matrix (i.e., a data structure containing multi-dimensional features that comprehensively reflects the operating status of the equipment).

[0032] Step S30: Based on the multi-dimensional fault feature matrix, the fault location of the weighing sensor is performed through a deep learning model to determine fault detection information.

[0033] It should be noted that deep learning models can automatically extract features from data through multi-layer nonlinear transformations and can handle complex nonlinear relationships. In this embodiment, a deep learning model is used to extract spatial and temporal features from a multidimensional fault feature matrix and accurately classify and locate faults using a fully connected neural network and attention mechanism.

[0034] This embodiment independently analyzes multi-source data from different sensors to obtain analyzed data in each dimension. This data is then correlated and fused to generate a multi-dimensional fault feature matrix. Based on the multi-dimensional fault feature matrix, a deep learning model is used to locate faults in the load cell and determine fault detection information. This application achieves accurate detection and location of load cell faults by independently analyzing multi-source data, correlatively fusing it, and locating faults using a deep learning model. This addresses the difficulty of existing technologies in coping with multi-source data fusion and anomaly identification under complex working conditions, significantly improving the comprehensiveness and accuracy of fault detection.

[0035] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 2, the step S10 may include steps S101 to S108: Step S101 : Based on a thermodynamic model and Fourier transform, the periodic fluctuation characteristics of the temperature data are extracted to obtain a correlation analysis result between the temperature anomaly and the equipment operating status.

[0036] It should be noted that multi-source data can include temperature data, humidity data, tilt angle data, casing integrity test data, inert gas concentration data, interference information, and abnormal data fluctuation information. These data reflect different aspects of the device's operating status. Furthermore, multi-source data can also include other types of data, such as vibration data, current data, sound data, and image data, which can also be used for fault detection and analysis. This embodiment does not limit the specific types of multi-source data; any data that can reflect the device's operating status can be included in the scope of multi-source data.

[0037] For example, to verify the correlation between multi-source data and load cell failures, temperature cycling tests combined with thermal imaging analysis can be used to verify the impact of temperature anomalies. Humidity effects can be assessed through constant humidity aging experiments combined with impedance spectroscopy monitoring. Helium mass spectrometry leak detection tests can be used to analyze the characteristics of sealing failures. Multi-parameter coupling experiments can also be designed to investigate interactions. The experimental design should include normal operating condition controls, cover the typical operating range, and focus on the differences in the impact of sudden and gradual changes in parameters. Time series analysis should be used to establish a correspondence between each parameter and the failure characteristics, thereby systematically verifying the correlation between data such as temperature, humidity, and inert gas concentration and sensor performance degradation.

[0038] It should also be noted that temperature data refers to the ambient or device surface temperature values collected by sensors and is used to monitor performance changes caused by thermal stress. Humidity data reflects the ambient moisture content and is used to assess the corrosive effects of humid environments on circuits and materials. Tilt angle data describes the angular deviation of the device relative to the horizontal plane and is used to detect structural deformation or installation anomalies. Casing integrity data determines the integrity of the device casing by monitoring the concentration changes of tracer gases such as helium in a sealed environment. Inert gas concentration data is used to detect gas leaks within the system. Interference information refers to external interference signals such as electromagnetic noise and mechanical vibration. Abnormal data fluctuation information refers to sudden changes or trend changes in the sensor output signal that deviate from the normal pattern and may indicate potential failures. These parameters together constitute a multi-dimensional monitoring system that provides fault warnings through feature extraction and correlation analysis.

[0039] It should be understood that in this step, a thermodynamic model is first used to simulate the temperature variation of the equipment under normal operating conditions to establish a temperature reference curve. Next, a Fourier transform is used to perform frequency domain analysis on the temperature data to extract its periodic fluctuation characteristics. By comparing the actual temperature data with the reference curve, abnormal temperature fluctuations can be identified and analyzed for their correlation with the equipment's operating status. For example, an abnormally high temperature increase may indicate a risk of overheating, while an abnormally low temperature may indicate a malfunction in the equipment's cooling system.

[0040] Step S102 : obtaining an evaluation result of the impact of humidity data on the performance of the load cell by identifying the long-term change trend of humidity data.

[0041] It's important to note that humidity data significantly impacts load cell performance, especially in high-humidity environments, where moisture can lead to measurement errors or even failure. Therefore, humidity data can be evaluated using an environmental humidity impact assessment algorithm to analyze long-term trends. For example, a sliding window statistical method can be used to calculate the mean and standard deviation of humidity data and identify abnormal humidity patterns. By assessing humidity's impact on sensor performance, it's possible to determine whether load cells are at risk of moisture or corrosion, providing early warning information for equipment maintenance.

[0042] In step S103, a Kalman filter method is used to analyze the dynamic offset characteristics of the tilt angle data to obtain an analysis result of the potential damage of the abnormal tilt angle to the equipment structure.

[0043] It's understood that tilt angle data reflects the stability of the equipment structure. Kalman filtering can be used to dynamically analyze tilt angle data, eliminating measurement noise and extracting its dynamic excursion characteristics. For example, Kalman filtering can predict the normal range of tilt angle variation and identify abnormal excursions outside this range. Abnormal tilt angles may indicate a risk of looseness or deformation in the equipment structure, potentially jeopardizing its safe operation.

[0044] Step S104 : monitoring the helium concentration data in the sealed environment in real time, and generating a shell integrity detection result based on the helium concentration data.

[0045] It's important to note that helium, as an inert gas, is often used to detect leaks in sealed environments due to its high diffusivity and low background concentration. Helium sensors can collect real-time changes in helium concentration within a sealed environment and transmit the data to an analysis system. Abnormally high helium concentrations may indicate a risk of damage to the sealed enclosure or leakage.

[0046] In addition, this embodiment does not limit the type of inert gas used. In addition to helium, other inert gases such as neon, argon, krypton or xenon can also be used. These inert gases also have the characteristics of stable chemical properties and not easy to react with other substances, and are suitable for detecting the integrity of sealed environments. The specific choice of inert gas can be flexibly adjusted according to factors such as actual application scenarios, detection sensitivity requirements and cost. For example, in some high-precision detection scenarios, helium with stronger diffusivity can be selected; and in cost-sensitive scenarios, argon with lower prices can be selected. The technical solution of the present application is applicable to a variety of inert gases and has wide applicability and flexibility.

[0047] As an implementation method, the above-mentioned step S104 in this embodiment may include: real-time monitoring of helium concentration data in the sealed environment; using a pressure sensor to collect pressure fluctuation data of the sealed environment, and combining with a pressure fluctuation analysis algorithm to extract abnormal pressure fluctuation characteristics of the pressure fluctuation data; correlating the helium concentration data with the pressure fluctuation data to identify a coordinated change pattern of abnormal increase in helium concentration and sudden change in pressure fluctuation; and generating a shell integrity detection result by determining whether there is a shell damage event based on the coordinated change pattern.

[0048] It should be understood that in addition to helium concentration data, pressure fluctuation data of the sealed environment is also an important indicator of its integrity. In this step, the pressure fluctuation data of the sealed environment is collected in real time using a pressure sensor. Pressure fluctuation data may be affected by changes in the external environment or the operating status of the equipment, so it needs to be processed in conjunction with a pressure fluctuation analysis algorithm. For example, a sliding window statistical method is used to calculate the mean and standard deviation of pressure fluctuations and identify abnormal pressure fluctuation characteristics. Abnormal pressure fluctuations may indicate the risk of shell damage or leakage in the sealed environment.

[0049] There is a potential correlation between helium concentration data and pressure fluctuation data. Correlation analysis methods can be used to integrate these data and identify patterns of synergistic changes between them. For example, a correlation analysis algorithm can be used to calculate the correlation coefficient between helium concentration and pressure fluctuations, identifying patterns of synergistic changes between abnormal increases in helium concentration and sudden changes in pressure fluctuations. This synergistic change pattern is crucial for determining whether a sealed environment contains damage or leakage. Correlation analysis can significantly improve the accuracy and reliability of enclosure integrity testing.

[0050] After identifying the synergistic pattern of an abnormally high helium concentration and a sudden pressure fluctuation, further investigation is needed to determine whether a housing breach event exists. By setting a threshold or using a machine learning model to classify the synergistic pattern, the presence of a housing breach risk in the sealed environment can be determined. For example, if an abnormally high helium concentration and a sudden pressure fluctuation occur simultaneously, it can be determined that a housing breach event exists in the sealed environment. Based on this determination, a housing integrity test result is generated.

[0051] Step S105 , using a gas leakage model to analyze the gradient change of the inert gas concentration data to obtain a gas leakage fault location result, wherein the gas leakage model is a model generated based on the diffusion law and concentration gradient change of the inert gas inside the equipment.

[0052] It's understandable that using a gas leakage model (i.e., a model generated based on the diffusion patterns and concentration gradients of inert gases within the device) to analyze the gradient of inert gas concentration data can help pinpoint the source of a gas leak. For example, by calculating areas with high concentration gradients, the source can be located.

[0053] For example, the establishment of a gas leakage model is based on the physical laws of inert gas diffusion. Its core algorithm steps can be as follows: First, a basic diffusion equation is established based on Fick's diffusion law. Then, the finite element method is used to discretize the solution region, combining the internal structural characteristics of the device, and transforming the continuous diffusion problem into a node concentration calculation. Then, measured boundary conditions (including leak source intensity, ambient pressure, etc.) are introduced to construct the constraint equation. During the concentration gradient calculation stage, the five-point difference method is used to solve the concentration gradient at each spatial location. Finally, the leak location is accurately located by setting a leak judgment threshold and a pattern recognition algorithm (such as an SVM classifier). By coupling the diffusion equation with the device topology constraints, this model can accurately reflect the dynamic diffusion process of inert gases in complex structures. The value of the diffusion coefficient is determined through calibration experiments, and the influence of environmental factors such as temperature and pressure is taken into account.

[0054] Step S106 , separating the noise signal from the interference information by using an adaptive filtering algorithm to obtain effective signal data after interference elimination.

[0055] In real-world operating environments, load cell data is often affected by various interference signals. In this step, an adaptive filtering algorithm processes this interference information, separating the noise signal while retaining the valid signal. For example, the adaptive filter's dynamic adjustment capability allows real-time filtering of high-frequency noise and low-frequency interference, extracting valid signal data that reflects the device's true status.

[0056] Step S107 , based on the time series decomposition algorithm, by identifying the mutation points in the abnormal fluctuation information of the data, the correlation analysis results between the abnormal fluctuation and the fault are obtained.

[0057] It's understandable that abnormal data fluctuations can be analyzed using a time series decomposition algorithm to identify sudden changes. For example, by decomposing the trend, cycle, and residual terms of time series data, the occurrence and magnitude of abnormal fluctuations can be precisely located. By analyzing the correlation between abnormal fluctuations and equipment failures, the type and severity of the failure can be determined.

[0058] Step S108 , combining the correlation analysis result, the impact assessment result, the potential hazard analysis result, the shell integrity detection result, the positioning result, the valid signal data and the correlation analysis result to obtain the analyzed data of each dimension.

[0059] After completing the above steps, the temperature anomaly correlation analysis results, humidity impact assessment results, tilt angle potential hazard analysis results, shell integrity detection results, gas leakage location results, effective signal data after interference elimination, and abnormal fluctuation correlation analysis results are integrated to generate post-analysis data in each dimension.

[0060] This embodiment refines the independent analysis steps for multi-source data, including processing of temperature, humidity, tilt angle, housing integrity detection, inert gas concentration, interference information, and abnormal data fluctuation information. Through technical means such as thermodynamic models, Kalman filtering, and gas leakage models, the characteristics of various types of data are extracted and analysis results are generated. This embodiment can accurately identify the characteristics of various sensor failures, avoiding the limitations of single data source analysis. In addition, combined with helium filling technology and gas leakage models, it can effectively detect housing damage and gas leakage events, improving the coverage and reliability of fault detection.

[0061] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 3 , the step S30 may include steps S301 to S303: Step S301: input the multi-dimensional fault feature matrix into a pre-trained deep learning model to obtain spatial features and temporal features of the fault features.

[0062] It's understandable that the multidimensional fault feature matrix is fed into a pre-trained deep learning model to extract the spatial and temporal characteristics of the fault features. Deep learning models can include modules such as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks. For example, CNNs use convolutional and pooling layers to extract the spatial distribution of fault features, capturing failure modes at different locations on the device. LSTMs use time series analysis to extract the temporal evolution of fault features and identify dynamic trends. By extracting both spatial and temporal features, deep learning models can comprehensively capture the multi-dimensional nature of faults.

[0063] For example, for load cell fault detection, a hybrid neural network architecture combining CNN and LSTM can be used. The CNN layer contains three convolutional layers (with kernel sizes of 32×5, 64×3, and 128×3, respectively, and uses the Linear Rectification function (ReLU) and batch normalization) to extract spatial features. The LSTM layer uses a two-layer structure (128 units per layer) to capture temporal dependencies. This model is trained using the cross-entropy loss function and the Adam optimizer (Adaptive Moment Estimation) with a learning rate of 0.001. The dataset is derived from multi-source sensor data (temperature, humidity, vibration, etc.) collected from industrial sites and the corresponding fault labels. Compared to existing time series feature extraction frameworks, this architecture incorporates three optimizations tailored to the characteristics of weighing sensor data: 1) a dilated convolutional layer is added to the CNN to expand the receptive field and better capture multi-scale fault features; 2) an attention mechanism is introduced after the LSTM layer to highlight fault signs at critical time points; and 3) a specific data augmentation strategy is designed to improve model robustness by adding Gaussian perturbations tailored to sensor noise characteristics. These improvements enable the model to better adapt to the small sample size and noisy data characteristics of weighing sensor fault detection while maintaining the advantages of the original architecture.

[0064] Step S302: Generate a comprehensive fault feature based on the spatial feature and the temporal feature.

[0065] After obtaining the spatial features and temporal features, these features need to be fused to generate comprehensive fault features. For example, the spatial features and temporal features can be concatenated or weighted summed through a feature fusion layer to generate comprehensive fault features.

[0066] Step S303: Based on the comprehensive fault characteristics, output the fault detection information of the weighing sensor according to the fully connected neural network.

[0067] In this step, the comprehensive fault features are input into the fully connected neural network, and the high-order features of the fault can be extracted through multi-layer nonlinear transformation to obtain fault detection information.

[0068] As an implementation method, the above-mentioned step S303 in this embodiment may include: inputting the comprehensive fault features into a fully connected neural network, and extracting high-order features of the fault through multi-layer nonlinear transformation; based on the high-order features, using the attention mechanism to dynamically assign the contribution weights of different features to fault location to obtain a weight distribution result; according to the weight distribution result, outputting the fault type through a fault classifier, and combining the location information in the high-order features to generate a heat map of the fault location; based on the fault type and the heat map, outputting the fault detection information of the weighing sensor, wherein the fault detection information includes the fault type, fault location and fault severity.

[0069] It's easy to understand that comprehensive fault features are input into a fully connected neural network, where high-level fault features are extracted through multiple layers of nonlinear transformations. Each layer of the fully connected neural network applies nonlinear mapping to the input features, progressively extracting deeper fault patterns. For example, the hidden layer's activation function (such as ReLU) applies nonlinear transformations to the features, capturing the complex relationships within the fault. High-level features are a further abstraction and refinement of comprehensive fault features, more accurately reflecting the essential attributes of the fault.

[0070] Although high-order features contain rich fault information, the contribution of different features to fault location may vary. In this step, the attention mechanism is used to dynamically assign weights to the contributions of different features to fault location. The attention mechanism dynamically adjusts the weight distribution of features by calculating the importance scores of the features. For example, the attention mechanism assigns higher weights to certain key areas in spatial features or certain key time points in temporal features, thereby enhancing the significance of these features in fault location. After obtaining the weight distribution results, the weighted high-order features are input into the fault classifier, and the fault type is output. The fault classifier can use the softmax function to classify features and determine the specific type of fault. At the same time, combined with the location information in the high-order features, a heat map of the fault location is generated. The heat map intuitively displays the distribution of fault locations through color gradients. For example, red areas indicate locations with a higher probability of faults, and blue areas indicate locations with a lower probability of faults.

[0071] After fault classification and heat map generation, the fault type and heat map information are integrated to output load cell fault detection information. This information includes the fault type (e.g., overheating, leakage, tilt), fault location (displayed via a heat map), and fault severity (e.g., mild, moderate, severe).

[0072] This embodiment extracts spatial and temporal features from a multidimensional fault feature matrix to comprehensively capture multidimensional fault information and improve the accuracy of fault location. The application of a fully connected neural network further enhances the ability to classify and locate faults, enabling rapid output of fault type and location information. Furthermore, the application of an attention mechanism dynamically adjusts the weights of different features, enhancing the significance of key features. The fault location heat map further enhances the visualization of fault detection, making fault information more intuitive.

[0073] As an implementation manner, step S20 may include: Based on a data alignment algorithm, performing time synchronization and spatial alignment on the analyzed data to obtain aligned data; A multi-dimensional fault feature matrix is constructed by assigning weights to the aligned data.

[0074] It should be understood that the analyzed data is characteristic data obtained after independent analysis of multi-source data collected from different sensors, such as temperature anomalies, humidity change trends, etc. However, these data may be inconsistent in time and space, and direct fusion may lead to information distortion. In this step, the analyzed data are time synchronized and spatially aligned through the data alignment algorithm. The purpose of time synchronization is to unify the time series of different data sources onto the same time axis, such as using interpolation methods to fill in missing time points or using dynamic time warping algorithms to eliminate time offsets. The purpose of spatial alignment is to map the spatial position information of different data sources into a unified coordinate system. Through time synchronization and spatial alignment, aligned data are obtained, which are consistent in time and space.

[0075] Although aligned data resolves temporal and spatial inconsistencies, the contribution of different data to fault detection may vary. In this step, weights are assigned to the aligned data to construct a multidimensional fault feature matrix. The weight assignment can be dynamically adjusted based on historical fault data and real-time data analysis results. For example, a machine learning model can be used to calculate the importance score of each type of data for fault detection and assign it a corresponding weight. A multidimensional fault feature matrix is a data structure containing multidimensional features that can comprehensively reflect the operating status of the equipment. For example, each row of the matrix represents a time point, each column represents a feature dimension, and each element in the matrix represents the weight value of the corresponding feature at a specific time point.

[0076] As an implementation method, in this embodiment, the above-mentioned data alignment algorithm is based on which the analyzed data is time synchronized and spatially aligned, and the steps of obtaining the aligned data may include: based on the data alignment algorithm, time aligning the analyzed data to obtain preliminary time aligned data; eliminating the time offset of the preliminary time aligned data through a dynamic time warping algorithm to obtain time synchronization data; spatially aligning the time synchronization data to obtain preliminary spatial alignment data; based on the topological relationship of the device structure, mapping the preliminary spatial alignment data to a unified fault coordinate system to obtain spatial alignment data; combining the time synchronization data and the spatial alignment data to obtain aligned data.

[0077] It will be appreciated that in this step, the analyzed data is first time-aligned using a data alignment algorithm. For example, interpolation can be used to pad low-sampling frequency data onto a high-sampling frequency timeline, or timestamp alignment can be used to align the start times of different data sources. After time alignment, preliminary time-aligned data is obtained, which is consistent on the timeline.

[0078] While the initial time-aligned data resolves inconsistencies along the timeline, time skew may still exist. This means that data from different data sources at the same point in time may be out of sync due to transmission delays or differences in processing time. In this step, the time skew of the initial time-aligned data is eliminated using the Dynamic Time Warping (DTW) algorithm. The DTW algorithm minimizes time skew by calculating the optimal matching path between different time series. For example, for temperature and humidity data, the DTW algorithm can find the optimal temporal alignment between the two, eliminating time skew.

[0079] Then, the time-synchronized data is spatially aligned. For example, the spatial positions of different sensors are mapped to the same reference coordinate system through coordinate transformation, or sparse spatial data is padded into a unified grid through interpolation methods. After spatial alignment, preliminary spatial alignment data is obtained. Next, based on the topological relationship of the device structure, the preliminary spatial alignment data is mapped to a unified fault coordinate system. For example, a unified fault coordinate system is established with the core fault area of the device structure as the origin, and the spatial position information of the sensors is converted into this coordinate system based on the device topology. After mapping, spatial alignment data is obtained, and these data are consistent in the unified fault coordinate system.

[0080] After completing time synchronization and spatial alignment, the time-synchronized data and spatially aligned data are combined to generate aligned data. Aligned data is consistent across time and space, fully reflecting the multi-dimensional characteristics of the device's operating status. For example, each row of the aligned data represents a point in time, each column represents a feature dimension, and each element in the matrix represents the value of the corresponding feature at a specific time and spatial location.

[0081] This embodiment solves the problem of temporal and spatial inconsistency of multi-source data through time synchronization and spatial alignment processing, ensuring the accuracy of data fusion. The weight distribution further optimizes the construction of the multi-dimensional fault feature matrix and improves the input quality of the fault detection model. The dynamic time warping algorithm effectively eliminates the time offset problem of multi-source data and ensures the accuracy of time synchronization. The spatial alignment based on the device topology relationship further improves the spatial consistency of the data.

[0082] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the fault detection method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0083] This application also provides a fault detection device, please refer to Figure 4 , the fault detection device comprises: The data processing module 10 is used to independently analyze multi-source data from different sensors to obtain analyzed data in various dimensions; The data fusion module 20 is used to perform correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix; The information acquisition module 30 is used to locate the fault of the weighing sensor based on the multi-dimensional fault feature matrix through a deep learning model to determine the fault detection information.

[0084] The fault detection device provided in this application, which utilizes the fault detection method described in the aforementioned embodiments, can address the technical issues of existing load cell fault detection technologies, which rely heavily on a single data source and struggle to address multi-source data fusion and anomaly identification under complex operating conditions. Compared to the prior art, the beneficial effects of the fault detection device provided in this application are the same as those of the fault detection method described in the aforementioned embodiments. Other technical features of the fault detection device are the same as those disclosed in the aforementioned embodiments and are not further elaborated here.

[0085] The present application provides a fault detection device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the fault detection method in the above-mentioned embodiment one.

[0086] Reference below Figure 5 , which shows a schematic diagram of the structure of a fault detection device suitable for implementing the embodiments of the present application. The fault detection device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Devices), PMPs (Portable Media Players), in-vehicle terminals (such as in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The fault detection device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0087] like Figure 5 As shown, the fault detection device may include a processing device 1001 (e.g., a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory 1002 or programs loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the fault detection device. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems may be connected to the input / output interface 1006: an input device 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; a storage device 1003 including, for example, a magnetic tape or hard disk; and a communication device 1009. Communication device 1009 can allow the fault detection device to communicate with other devices wirelessly or wired to exchange data. Although the figure shows a fault detection device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.

[0088] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0089] The fault detection device provided in this application, utilizing the fault detection method described in the aforementioned embodiment, can address the technical issues of existing load cell fault detection technologies, which often rely on a single data source and struggle to address multi-source data fusion and anomaly identification under complex operating conditions. Compared to the prior art, the beneficial effects of the fault detection device provided in this application are the same as those of the fault detection method described in the aforementioned embodiment. The other technical features of this fault detection device are the same as those disclosed in the aforementioned embodiment and are not further elaborated here.

[0090] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0091] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0092] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, a computer program) stored thereon, wherein the computer-readable program instructions are used to execute the fault detection method in the above embodiment.

[0093] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0094] The computer-readable storage medium may be included in the fault detection device; or may exist independently without being assembled into the fault detection device.

[0095] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the fault detection device, the fault detection device enables the fault detection device to: independently analyze multi-source data from different sensors, obtain analyzed data of each dimension, perform correlation fusion on the analyzed data, generate a multi-dimensional fault feature matrix, and based on the multi-dimensional fault feature matrix, locate the fault of the weighing sensor through a deep learning model to determine the fault detection information.

[0096] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0097] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0098] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0099] The computer-readable storage medium provided in this application stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned fault detection method. This computer-readable storage medium can address the technical issues of existing load cell fault detection technologies, which often rely on a single data source and struggle to address multi-source data fusion and anomaly identification under complex operating conditions. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are similar to those of the fault detection method provided in the aforementioned embodiments and are not further elaborated here.

[0100] The present application also provides a computer program product, comprising a computer program, which implements the steps of the above-mentioned fault detection method when executed by a processor.

[0101] The computer program product provided in this application can address the technical issues of existing load cell fault detection technologies, which often rely on a single data source and struggle to address multi-source data fusion and anomaly identification under complex operating conditions. Compared to existing technologies, the beneficial effects of the computer program product provided in this application are similar to those of the fault detection methods provided in the aforementioned embodiments and are not further elaborated here.

[0102] The above descriptions are only some embodiments of the present application and do not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A fault detection method, characterized in that: The method comprises the following steps: Independently analyze multi-source data from different sensors to obtain analyzed data in various dimensions; Performing correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix; Based on the multi-dimensional fault feature matrix, the fault location of the weighing sensor is performed through a deep learning model to determine the fault detection information.

2. The fault detection method according to claim 1, wherein: The multi-source data includes temperature data, humidity data, tilt angle data, housing integrity detection data, inert gas concentration data, interference information, and abnormal data fluctuation information. The step of independently analyzing the multi-source data from different sensors to obtain analyzed data in each dimension includes: Based on thermodynamic models and Fourier transform, the correlation analysis results between temperature anomalies and equipment operating status are obtained by extracting the periodic fluctuation characteristics of temperature data; By identifying the long-term change trend of humidity data, we can evaluate the impact of humidity data on the performance of load cells. The Kalman filter method is used to analyze the dynamic offset characteristics of the tilt angle data to obtain the potential damage analysis results of the abnormal tilt angle to the equipment structure; monitoring helium concentration data in the sealed environment in real time, and generating a shell integrity test result based on the helium concentration data; Utilizing a gas leakage model, by analyzing the gradient change of the inert gas concentration data, the gas leakage fault location result is obtained, wherein the gas leakage model is generated based on the diffusion law and concentration gradient change of the inert gas inside the equipment; By using adaptive filtering algorithm to separate the noise signal from the interference information, the effective signal data after the interference is eliminated is obtained; Based on the time series decomposition algorithm, by identifying the mutation points in the abnormal fluctuation information of the data, the correlation analysis results between abnormal fluctuation and fault are obtained; The analyzed data of each dimension is obtained by combining the correlation analysis results, the impact assessment results, the potential hazard analysis results, the shell integrity detection results, the positioning results, the valid signal data and the correlation analysis results.

3. The fault detection method according to claim 2, wherein: The step of real-time monitoring of helium concentration data in the sealed environment and generating a housing integrity test result based on the helium concentration data includes: Real-time monitoring of helium concentration data in a sealed environment; Use pressure sensors to collect pressure fluctuation data in sealed environments, and combine them with pressure fluctuation analysis algorithms to extract abnormal pressure fluctuation characteristics of the pressure fluctuation data; Correlation analysis is performed on the helium concentration data and the pressure fluctuation data to identify a coordinated change pattern of abnormal increase in helium concentration and sudden change in pressure fluctuation; According to the coordinated change pattern, a shell integrity detection result is generated by determining whether a shell damage event occurs.

4. The fault detection method according to any one of claims 1 to 3, characterized in that: The step of performing correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix includes: Based on a data alignment algorithm, performing time synchronization and spatial alignment on the analyzed data to obtain aligned data; A multi-dimensional fault feature matrix is constructed by assigning weights to the aligned data.

5. The fault detection method according to claim 4, wherein: The step of performing time synchronization and spatial alignment on the analyzed data based on a data alignment algorithm to obtain aligned data includes: Based on a data alignment algorithm, time-aligning the analyzed data to obtain preliminary time-aligned data; Eliminating the time offset of the preliminary time alignment data by a dynamic time warping algorithm to obtain time synchronization data; performing spatial alignment on the time synchronization data to obtain preliminary spatial alignment data; Based on the topological relationship of the device structure, the preliminary spatial alignment data is mapped to a unified fault coordinate system to obtain spatial alignment data; The time synchronization data and the spatial alignment data are combined to obtain aligned data.

6. The fault detection method according to any one of claims 1 to 3, characterized in that: The step of locating the fault of the weighing sensor by using a deep learning model based on the multi-dimensional fault feature matrix and determining the fault detection information includes: Inputting the multidimensional fault feature matrix into a pre-trained deep learning model to obtain spatial features and temporal features of the fault features; generating a comprehensive fault feature based on the spatial feature and the temporal feature; Based on the comprehensive fault characteristics, fault detection information of the weighing sensor is output according to a fully connected neural network.

7. The fault detection method according to claim 6, wherein: The step of outputting fault detection information of the weighing sensor according to a fully connected neural network based on the comprehensive fault characteristics includes: Inputting the comprehensive fault features into a fully connected neural network, and extracting high-order features of the fault through multi-layer nonlinear transformation; Based on the high-order features, the attention mechanism is used to dynamically assign the contribution weights of different features to fault localization to obtain a weight distribution result; According to the weight distribution result, the fault type is output by the fault classifier, and the location information in the high-order features is combined to generate a heat map of the fault location; Based on the fault type and the thermal map, fault detection information of the weighing sensor is output, wherein the fault detection information includes the fault type, the fault location, and the fault severity.

8. A fault detection device, characterized in that: The fault detection device comprises: The data processing module is used to independently analyze multi-source data from different sensors to obtain analyzed data in various dimensions; A data fusion module is used to perform correlation fusion on the analyzed data to generate a multi-dimensional fault feature matrix; The information acquisition module is used to locate the fault of the weighing sensor based on the multi-dimensional fault feature matrix through a deep learning model to determine the fault detection information.

9. A fault detection device, characterized in that: The fault detection device includes: a memory, a processor, and a fault detection program stored in the memory and executable on the processor. When the fault detection program is executed by the processor, the fault detection method according to any one of claims 1 to 7 is implemented.

10. A storage medium, characterized in that: The storage medium stores a fault detection program, which, when executed by a processor, implements the fault detection method according to any one of claims 1 to 7.

Images