Digital light source controller fault diagnosis method and device

Abstract

The present invention relates to a digital light source controller fault diagnosis method and device, the method comprises obtaining the original sensor data of the digital light source controller, and performing multi-dimensional analysis and extraction to obtain corresponding multi-dimensional sensor data; performing time domain and frequency domain feature extraction on the multi-dimensional sensor data to obtain corresponding sensor feature data; performing fault feature analysis on the sensor feature data to obtain corresponding fault type probability distribution; performing dynamic threshold judgment on the fault type probability distribution to obtain corresponding preliminary fault analysis results; performing sampling period feature fault diagnosis on the preliminary fault analysis results to obtain corresponding fault diagnosis results; performing hierarchical processing on the fault diagnosis results, and performing scheme analysis to obtain corresponding fault handling schemes. The present invention can combine fault feature analysis to realize accurate identification and classification of various controller faults.

Description

Technical Field

[0001] The present invention relates to the technical field of digital light source controllers, and in particular to a digital light source controller fault diagnosis method and device. Background Art

[0002] As the core control equipment in modern lighting systems, digital light source controllers have been widely used in industrial production, commercial lighting, and intelligent buildings. With the continuous improvement of intelligent manufacturing and precise control requirements, how to timely and accurately diagnose the operating status and fault conditions of digital light source controllers to ensure the stable operation and control effect of lighting systems has become an important topic of concern in the industry. Existing fault diagnosis methods are often limited to single-dimensional data analysis, such as focusing only on the current parameters or temperature changes of the controller, while ignoring the collaborative analysis of multi-dimensional sensor data and the dynamic characteristics of fault evolution. This single diagnostic method is prone to cause one-sidedness and lag in fault judgment, which in turn affects the reliability and service life of the lighting system. Summary of the invention

[0003] The main purpose of the present invention is to provide a digital light source controller fault diagnosis method and device, which can combine fault feature analysis to achieve accurate identification and classification of various controller faults.

[0004] To achieve the above object, the present invention provides a digital light source controller fault diagnosis method, comprising: Acquire the original sensor data of the digital light source controller, and perform multi-dimensional analysis and extraction to obtain corresponding multi-dimensional sensor data; Extracting time-domain and frequency-domain features from the multidimensional sensor data to obtain corresponding sensor feature data; Performing fault feature analysis on the sensor feature data to obtain a corresponding fault type probability distribution; Performing dynamic threshold judgment on the probability distribution of the fault type to obtain corresponding preliminary fault analysis results; Perform sampling period characteristic fault diagnosis on the preliminary fault analysis result to obtain a corresponding fault diagnosis result; The fault diagnosis results are graded and analyzed to obtain corresponding fault handling solutions.

[0005] Furthermore, the original sensor data of the digital light source controller is obtained, and multi-dimensional analysis and extraction are performed to obtain corresponding multi-dimensional sensor data, including: Performing data collection on the digital light source controller to obtain original sensor data; Performing frequency band matrix decomposition on the original sensor data to obtain coefficient matrices of different frequency bands; Reconstructing the coefficient matrix to obtain reconstructed signal data; Performing Fourier transform on the reconstructed signal data to obtain frequency spectrum feature data; Performing peak distribution analysis on the frequency spectrum feature data to obtain multi-dimensional sensing features; The multi-dimensional sensing features are subjected to data fusion to obtain the multi-dimensional sensing data.

[0006] Furthermore, the extracting time domain and frequency domain features of the multi-dimensional sensing data to obtain corresponding sensing feature data includes: Performing empirical mode decomposition on the multidimensional sensing data to obtain a light source wave mode function sequence; Performing time domain feature calculation according to the light source wave mode function sequence to obtain a light source time-varying feature sequence; Performing Hilbert transform on the light source wave mode function sequence to obtain the light source instantaneous spectrum; Performing local linear embedding matrix construction on the instantaneous spectrum of the light source to obtain a light source state characteristic matrix; Performing bispectral analysis according to the instantaneous spectrum of the source to obtain a light source fault vector; The light source state feature matrix and the light source fault vector are sparsely autoencoded and fused to obtain the sensing feature data.

[0007] Furthermore, the performing fault feature analysis on the sensor feature data to obtain a corresponding fault type probability distribution includes: The sensing characteristic data is subjected to wavelet packet decomposition processing to obtain a multi-scale decomposition coefficient of a light source fault; Calculate the energy entropy eigenvalue according to the multi-scale decomposition coefficient of the light source fault to obtain the light source fault energy eigenvector; Performing adaptive cluster analysis on the light source fault energy feature vector to obtain a light source control feature cluster center; Constructing a Mahalanobis distance matrix according to the light source control feature clustering center to obtain light source controller feature distance data; Performing hierarchical analysis on the characteristic distance data of the light source controller to obtain the correlation degree of the light source fault characteristics; Perform fault mode identification according to the light source fault feature correlation to obtain a light source controller fault identification result; Establishing a radial basis function according to the light source controller fault identification result to obtain a light source controller fault probability density function; The fault type probability is calculated for the light source fault probability density function to obtain the fault type probability distribution.

[0008] Furthermore, the dynamic threshold judgment is performed on the fault type probability distribution to obtain the corresponding preliminary fault analysis result, including: Sorting the probability of fault types according to the fault type probability distribution to obtain a fault type probability sequence; Performing historical data comparison on the fault type probability sequence to obtain a fault type probability deviation value; Performing multi-level threshold division on the fault type probability deviation value to obtain a fault type threshold interval; Performing classification statistics of fault types according to the fault type threshold interval to obtain a fault type frequency distribution; Performing trend prediction on the frequency distribution of the fault types to obtain a predicted value of the fault development trend; Calculating the matching degree of the fault development trend prediction value according to the preset fault judgment standard to obtain the fault type matching degree; Fault analysis is performed on the fault type matching degree and the fault type frequency distribution to obtain the preliminary fault analysis result.

[0009] Furthermore, the sampling period characteristic fault diagnosis is performed on the preliminary fault analysis result to obtain the corresponding fault diagnosis result, including: Performing period segmentation on the preliminary fault analysis result to obtain a corresponding periodic data sequence; Perform fault state matching analysis according to the periodic data sequence to obtain corresponding state transition information; Processing the state transition information in segments to obtain a corresponding state interval sequence; Perform fault periodicity matching according to the state interval sequence to obtain a corresponding fault period vector; Performing multi-layer recursive analysis on the fault cycle vector to obtain a corresponding fault state sequence; Performing type probability calculation according to the fault state sequence to obtain corresponding fault diagnosis confidence; Fault diagnosis is performed on the fault diagnosis confidence to obtain the fault diagnosis result.

[0010] Furthermore, the fault diagnosis results are graded and analyzed to obtain corresponding fault handling solutions, including: Performing fault classification on the fault diagnosis results to obtain fault classification data including fault impact degree, fault frequency, fault duration, fault severity, repair difficulty and fault repair time; Perform a multi-level comprehensive evaluation based on the dimensional fault classification data to obtain fault level evaluation information; Performing fault priority sorting calculation on the fault level assessment information to obtain a fault processing priority sequence; Allocate repair resources according to the fault handling priority sequence to obtain a fault repair resource configuration plan; Performing repair arrangement analysis according to the fault repair resource configuration plan to obtain a fault repair execution sequence; The fault repair resource configuration scheme and the fault repair execution sequence are integrated to obtain the fault handling scheme.

[0011] The present invention further provides a digital light source controller fault diagnosis device, which is applied to any one of the digital light source controller fault diagnosis methods described above, comprising: An acquisition module, which is used to acquire original sensor data of the digital light source controller and perform multi-dimensional analysis and extraction to obtain corresponding multi-dimensional sensor data; An analysis module, the analysis module is used to extract time domain and frequency domain features of the multi-dimensional sensor data to obtain corresponding sensor feature data; An association module, the association module is used to perform fault feature analysis on the sensor feature data to obtain a corresponding fault type probability distribution; A processing module, the processing module is used to perform dynamic threshold judgment on the probability distribution of the fault type to obtain a corresponding preliminary fault analysis result; A control module, the control module is used to perform sampling period characteristic fault diagnosis on the preliminary fault analysis result to obtain a corresponding fault diagnosis result; An execution module is used to perform hierarchical processing on the fault diagnosis results and perform solution analysis to obtain a corresponding fault processing solution.

[0012] A digital light source controller fault diagnosis method and device provided by the present invention has the following beneficial effects: By performing multi-dimensional analysis, extraction and feature analysis on the original sensor data of the digital light source controller, the operating status characteristics of the controller can be fully captured, the accuracy of fault diagnosis is significantly improved, and reliable data support is provided for fault prevention and processing. By extracting the time domain and frequency domain features of the sensor data and combining the fault feature analysis, the various faults of the controller can be accurately identified and classified, which helps to quickly locate the cause of the fault and reduce the system downtime. Dynamic threshold judgment based on the probability distribution of fault types can ensure that the system can detect potential faults in time under different operating conditions, reduce the misjudgment rate, and improve the reliability of the diagnosis system. Through sampling period feature fault diagnosis and hierarchical processing, a more reasonable fault handling plan is formulated, and the classification processing of faults is realized through hierarchical response strategies, thereby effectively improving the maintenance efficiency of the system and extending the service life of the equipment. At the same time, by performing hierarchical analysis on the fault diagnosis results, the processing strategy can be flexibly adjusted according to the severity and type of different faults, making the system more adaptable to complex and changeable application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a flow chart of a digital light source controller fault diagnosis method provided by the present invention; Figure 2 The present invention provides a structure diagram of a digital light source controller fault diagnosis device.

[0014] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0017] Reference Figure 1 As shown, the present invention provides 1. a digital light source controller fault diagnosis method, characterized in that it includes: Step S1: acquiring original sensor data of the digital light source controller, and performing multi-dimensional analysis and extraction to obtain corresponding multi-dimensional sensor data; Step S2: extracting time domain and frequency domain features from the multi-dimensional sensor data to obtain corresponding sensor feature data; Step S3: performing fault feature analysis on the sensor feature data to obtain corresponding fault type probability distribution; Step S4: Perform dynamic threshold judgment on the probability distribution of fault types to obtain corresponding preliminary fault analysis results; Step S5: Perform sampling period characteristic fault diagnosis on the preliminary fault analysis results to obtain corresponding fault diagnosis results; Step S6: classify the fault diagnosis results and perform solution analysis to obtain a corresponding fault handling solution.

[0018] Based on the above steps, the detailed process is as follows: Step S1: The digital light source controller will generate a large amount of raw sensor data during operation. These data sources include multiple sensor collection points such as voltage sensors, current sensors, temperature sensors, and light intensity sensors. First, a complete data acquisition system needs to be established to ensure that these raw data can be acquired in real time and accurately. The data acquisition system should have anti-interference capabilities and ensure that the data sampling frequency meets the fault diagnosis requirements. After obtaining the raw data, data preprocessing is required, including operations such as removing outliers, eliminating noise, and supplementing missing values ​​to ensure data quality. Subsequently, the preprocessed data is reduced in dimension and feature extracted by a multidimensional analysis method. This process needs to consider the correlation between different sensor data and retain the data features with the most fault diagnosis value. The multidimensional sensor data finally obtained should be able to better reflect the operating status of the equipment and provide a reliable data basis for subsequent feature extraction. The output of this step directly affects the effect of subsequent feature extraction, so special attention should be paid to the integrity and accuracy of the data.

[0019] Step S2: Based on the obtained multi-dimensional sensor data, it is necessary to extract features in both the time domain and the frequency domain. In the time domain analysis, the main focus is on the statistical characteristics of the data, including statistical quantities such as mean, standard deviation, peak, skewness, and kurtosis. These characteristics can reflect the overall trend and fluctuation of the equipment's operating status. At the same time, it is necessary to analyze the time series change characteristics of the data, including dynamic characteristics such as rise time, fall time, and steady-state error. These characteristics are of great significance for discovering transient faults. In the frequency domain analysis, the time domain signal is converted to the frequency domain by Fourier transform and other methods to extract spectrum features, including main frequency components, harmonic components, frequency band energy distribution, etc. Frequency domain features can reflect the periodic faults and vibration characteristics of the equipment. Through the comprehensive analysis of the time domain and frequency domain, more comprehensive sensor feature data can be obtained. These feature data contain both the overall characteristics of the equipment's operating status and the local detailed characteristics, laying the foundation for subsequent fault feature analysis.

[0020] Step S3: After obtaining the sensor feature data, it is necessary to conduct an in-depth analysis of these features and establish a mapping relationship between the features and the fault types. Establish a fault feature library that contains the feature patterns of various typical faults. Through feature matching, the current sensor feature data is compared with the feature patterns in the fault feature library to calculate the similarity. In this process, the importance of different features is considered and the features are weighted. At the same time, the combination of faults is also considered, because multiple faults may exist at the same time in actual operation. Through feature matching and similarity calculation, the probability distribution of various possible fault types can be obtained. This probability distribution reflects the probability of various faults that may exist in the current equipment, and provides a basis for subsequent fault judgment. The analysis results of this step directly affect the accuracy of fault diagnosis, so it is necessary to fully consider various possible fault conditions and reasonably set the judgment criteria for feature matching.

[0021] Step S4: The probability distribution of fault types reflects the possibility of various faults. The dynamic threshold judgment mechanism determines the confidence level of the fault type by setting a reference threshold. The reference threshold consists of two parts: a fixed threshold and an adaptive threshold. The fixed threshold is formulated based on historical experience data, and the adaptive threshold is adjusted as the operating status of the equipment changes. During the judgment process, the probability values ​​of each fault type are compared, and the fault types above the threshold are marked as potential faults. The selection of thresholds is related to factors such as the operating environment, usage time, and load conditions of the equipment. In the judgment result, the time point, duration, severity, and other information of the fault are marked. This information constitutes the preliminary fault analysis results and provides basic data for the next sampling period characteristic fault diagnosis. The accuracy of the judgment result directly affects the final diagnosis effect. The system sorts the judgment results according to the credibility, and the fault types with high credibility are prioritized for the next analysis.

[0022] Step S5: Sampling cycle characteristic fault diagnosis combines the preliminary analysis results with the sampling cycle data. The system records the operating data of the equipment under different sampling cycles and analyzes the performance characteristics of the fault in different sampling cycles. During the diagnosis process, the system extracts the characteristic parameters of each sampling cycle and establishes a fault characteristic curve. By comparing the characteristic curves of different cycles, the development trend of the fault is identified. The changing trend of the characteristic curve shows the evolution process of the fault, including the initial stage, development stage and stable stage of the fault. The system determines the specific type and severity of the fault based on these stage characteristics. The fault diagnosis results include detailed information such as fault type, fault location, and fault severity. These diagnostic results have been verified over multiple sampling cycles and have high reliability, providing a basis for the formulation of subsequent fault handling plans.

[0023] Step S6: Fault classification processing classifies the diagnosis results according to the severity and urgency of the fault. The classification standards include multiple dimensions such as equipment performance impact, safety risks, and maintenance difficulty. Each level corresponds to different processing priorities and methods. Based on the classification results, the system matches the appropriate processing plan from the preset solution library. The processing plan includes specific contents such as maintenance steps, spare parts requirements, personnel configuration, and tool preparation. During the solution analysis process, the system evaluates the feasibility and effectiveness of each solution and selects the best solution. The processing plan also includes emergency measures and preventive measures to ensure that the fault is handled in a timely manner. After the final fault handling plan is optimized, a detailed execution document is formed to guide maintenance personnel to handle the fault. The execution effect of the processing plan will be recorded and used to update the solution library to optimize future fault handling processes.

[0024] A digital light source controller fault diagnosis method provided by the present invention can comprehensively capture the operating state characteristics of the controller by performing multi-dimensional analysis extraction and feature analysis on the original sensor data of the digital light source controller, significantly improve the accuracy of fault diagnosis, and provide reliable data support for fault prevention and processing. By extracting the time domain and frequency domain features of the sensor data and combining the fault feature analysis, accurate identification and classification of various controller faults are achieved, which helps to quickly locate the cause of the fault and reduce the system downtime. Dynamic threshold judgment based on the probability distribution of the fault type can ensure that the system can timely discover potential faults under different operating conditions, reduce the misjudgment rate, and improve the reliability of the diagnosis system. Through sampling period feature fault diagnosis and hierarchical processing, a more reasonable fault processing plan is formulated, and the classification processing of faults is realized through the hierarchical response strategy, thereby effectively improving the maintenance efficiency of the system and extending the service life of the equipment. At the same time, by performing hierarchical analysis on the fault diagnosis results, the processing strategy can be flexibly adjusted according to the severity and type of different faults, so that the system is more adaptable to complex and changeable application scenarios.

[0025] In one embodiment, the original sensor data of the digital light source controller is obtained, and multi-dimensional analysis and extraction are performed to obtain corresponding multi-dimensional sensor data, including: By collecting data from the digital light source controller, multiple sensors such as temperature sensors, voltage sensors, and current sensors are used to collect raw sensor data. The sampling frequency is set to 1kHz, the sampling time is 60s, and the number of sampling points is 60,000 points. The collected raw sensor data includes multi-dimensional data such as temperature data, voltage data, and current data.

[0026] The frequency band matrix of the collected raw sensor data is decomposed and constructed, and the wavelet packet decomposition method is used to decompose the raw sensor data into coefficient matrices of different frequency bands. The db4 wavelet basis function is selected, and the number of decomposition layers is 3, and the coefficient matrices of 8 frequency bands are obtained. The coefficient matrix of each frequency band contains the energy distribution information of the frequency band. The frequency band division range is: 0-62.5Hz, 62.5-125Hz, 125-187.5Hz, 187.5-250Hz, 250-312.5Hz, 312.5-375Hz, 375-437.5Hz, 437.5-500Hz.

[0027] Based on the obtained coefficient matrix, the wavelet packet reconstruction algorithm is used to reconstruct the coefficient matrix of each frequency band to obtain the reconstructed signal data. During the reconstruction process, the energy consistency between the reconstructed signal and the original signal is maintained, and the reconstruction error is controlled within 1%. The reconstructed signal data retains the characteristic information of the original signal in each frequency band.

[0028] The reconstructed signal data is subjected to Fourier transform, and the reconstructed signal is converted to the frequency domain using the fast Fourier transform algorithm to obtain the spectrum feature data. The number of sampling points of the Fourier transform is consistent with the number of sampling points of the original data, and the frequency resolution is 1 / 60Hz. The spectrum feature data contains the amplitude spectrum and phase spectrum information of the signal.

[0029] Based on the spectrum feature data, the peak distribution analysis is performed to extract the characteristic peak points in the spectrum. The peak extraction adopts the local maximum method, and the minimum interval between peak points is set to 10Hz, and the amplitude threshold is set to 20% of the root mean square value. The extracted peak features include multi-dimensional characteristic parameters such as peak frequency, peak amplitude, and peak bandwidth.

[0030] The extracted multi-dimensional sensing features are fused, and the principal component analysis method is used to reduce the dimension of the feature data. When fusion is performed, the principal components with a cumulative contribution rate of 95% are retained to form the final multi-dimensional sensing data. The fused data reduces data redundancy and improves the compactness of feature expression.

[0031] This embodiment realizes comprehensive monitoring of the operating status of the digital light source controller by adopting a multi-dimensional sensor data collection and analysis method. The frequency band matrix construction method based on wavelet packet decomposition enables the original sensor data to be accurately decomposed in different frequency bands, and the characteristic information of each frequency band is effectively extracted. The signal is restored by a reconstruction algorithm to ensure the energy consistency between the reconstructed signal and the original signal, and improve the accuracy of subsequent analysis. By combining Fourier transform and peak distribution analysis, the deep mining of spectral features is realized, providing a reliable feature basis for fault diagnosis. Finally, the principal component analysis method is used to fuse multi-dimensional features, which reduces data redundancy, improves the compactness of feature expression, and makes the fault diagnosis process more efficient and reliable. It not only improves the accuracy of fault diagnosis of digital light source controllers, but also reduces the computational complexity, providing effective technical support for real-time diagnosis in industrial sites.

[0032] In one embodiment, time domain and frequency domain feature extraction is performed on multi-dimensional sensor data to obtain corresponding sensor feature data, including: Empirical mode decomposition is the basic link of feature extraction. Multidimensional sensor data is decomposed into several intrinsic mode functions (IMFs), each of which represents an inherent oscillation mode. The decomposition process follows the following rules: the difference between the number of extreme points and the number of zero-crossing points of the signal does not exceed 1; at any moment, the average value of the envelope formed by the local maximum points and the envelope formed by the local minimum points is zero. Through repeated screening process, a complete IMF sequence, that is, the light source wave mode function sequence, is obtained.

[0033] Based on the acquired light source fluctuation mode function sequence, the time domain feature calculation focuses on extracting the statistical characteristics of the waveform. The calculation content includes time domain statistics such as mean, variance, peak factor, margin factor, etc. These statistics reflect the stability and fluctuation degree of the light source operation state. The extraction of time domain features follows the statistical principles, and forms the light source time-varying feature sequence through quantitative analysis of the waveform amplitude distribution, fluctuation law, etc.

[0034] The Hilbert transform converts the light source wave mode function sequence into the analytical signal space. The transformation process forms a complex signal with the original signal and its Hilbert transform, and then obtains the amplitude and phase information of the signal. The core of this step is to construct the instantaneous frequency spectrum of the signal, reflecting the dynamic change characteristics of the light source state over time.

[0035] The construction of the local linear embedding matrix is ​​based on the instantaneous spectrum of the light source. The process calculates the Euclidean distance between the spectrum data points, determines the set of neighboring points for each data point, and establishes a weight matrix. The optimization goal of the weight matrix is ​​to minimize the reconstruction error while maintaining the local geometric structure of the data. The final feature matrix reflects the low-dimensional essential characteristics of the light source state.

[0036] Bispectral analysis focuses on revealing the nonlinear characteristics in the signal. This analysis method calculates the third-order cumulant of the signal and obtains the bispectral estimation through Fourier transform. During the analysis process, the coupling relationship between frequency components is focused on, the characteristic frequency combination under the fault state of the light source is extracted, and finally the fault vector of the light source is formed.

[0037] Sparse autoencoder fusion is the last step of feature extraction. In this process, the light source state feature matrix and the light source fault vector are input into the autoencoder network. The encoder part realizes feature compression through multi-layer nonlinear transformation, and the decoder part restores the compressed features to the original dimension. Sparse constraints are introduced during the training process to ensure that the extracted features have good sparsity. The sensor feature data finally output realizes the effective fusion of time domain features and frequency domain features.

[0038] This embodiment uses empirical mode decomposition to process multidimensional sensor data, effectively extracts the inherent oscillation mode of light source fluctuations, and overcomes the strict requirements of traditional signal decomposition methods on signal stability. The feature extraction strategy based on Hilbert transform and local linear embedding realizes the accurate capture of the dynamic characteristics of the light source and significantly improves the accuracy of feature expression. The introduction of the bispectral analysis method deeply explores the nonlinear characteristics and frequency coupling information in the signal, providing a richer feature basis for fault diagnosis. The sparse autoencoder is used for feature fusion, which achieves dimensionality reduction and optimization of data while maintaining the integrity of time domain and frequency domain features, and improves the efficiency and accuracy of fault diagnosis. A complete feature extraction system has been established, which significantly enhances the reliability and accuracy of fault diagnosis through multi-level data analysis and fusion while ensuring the comprehensiveness of feature extraction.

[0039] In one embodiment, the fault feature analysis is performed on the sensor feature data to obtain the corresponding fault type probability distribution, including: When analyzing the fault characteristics of the digital light source controller fault diagnosis method, the sensor feature data is decomposed into four layers through the db4 wavelet packet. Based on the wavelet packet decomposition of the sensor feature data, the multi-scale decomposition coefficients of the light source fault including high frequency and low frequency are obtained. The db4 wavelet function is selected for wavelet packet decomposition because it has good local characteristics and orthogonality. The four-layer decomposition can effectively extract fault feature information.

[0040] On the basis of obtaining the multi-scale decomposition coefficients of the light source fault, the energy entropy characteristic value of each frequency band coefficient is calculated using the Shannon entropy principle. During the calculation process, the energy of each frequency band signal is normalized, and a probability distribution model is established. The energy entropy value of each frequency band is obtained by calculation, and then the energy characteristic vector of the light source fault is formed. This vector contains the energy distribution characteristics of the fault signal in different frequency bands.

[0041] The improved K-means algorithm is used to perform adaptive clustering analysis on the fault energy feature vector of the light source. During the clustering process, the density peak method is used to select the initial cluster center, and the optimal number of cluster centers is determined by calculating the local density and distance factor of the sample points. The clustering iterative optimization uses the Mahalanobis distance metric until the cluster center is stable and the light source control feature cluster center is formed.

[0042] Based on the obtained light source control feature cluster center, the Mahalanobis distance matrix is ​​constructed. The covariance matrix is ​​introduced in the distance calculation process, and the correlation between features is considered. By calculating the Mahalanobis distance from each feature point to the cluster center, the light source controller feature distance data matrix is ​​formed. This matrix reflects the distribution law of different fault type features.

[0043] A hierarchical analysis is performed on the feature distance data matrix, and the minimum variance method is used to construct a fault feature hierarchy tree. The optimal number of clustering levels is determined by calculating the ratio of the inter-class distance to the intra-class distance, and the correlation of the light source fault features is obtained. The calculation of the correlation is based on a weighted combination of Euclidean distance and cosine similarity.

[0044] Based on the obtained fault feature correlation, support vector machine is used for fault pattern recognition. By constructing the optimal classification hyperplane, accurate classification of fault types is achieved, and the fault recognition result of the light source controller is obtained. Gaussian kernel function is used in the recognition process, and the kernel parameters are determined by cross-validation.

[0045] Based on the fault identification results, a radial basis function network is constructed. The network structure adopts a three-layer architecture, the number of hidden layer nodes is determined by orthogonal least squares method, and the network training uses gradient descent method to optimize weight parameters, and finally the fault probability density function of the light source controller is obtained.

[0046] The fault probability density function is integrated to calculate the probability distribution of different fault types. The numerical integration method is used in the probability calculation process. The integration interval is determined according to the characteristic range of the fault type. Finally, the probability value of each type of fault is output to form the probability distribution of the fault type. This distribution reflects the probability of different fault types.

[0047] This embodiment uses db4 wavelet packets to perform 4-layer decomposition of sensor feature data, combined with energy entropy calculation and adaptive clustering analysis, to achieve accurate extraction of digital light source controller fault features, greatly improving the accuracy of fault diagnosis. Based on the Mahalanobis distance matrix and hierarchical analysis method, the feature data is deeply mined to effectively identify the correlation between fault features and enhance the reliability of the diagnosis results. The optimal classification hyperplane is constructed by support vector machine, and the probability density analysis of radial basis function network is combined to establish a complete fault probability assessment system, which realizes accurate identification and probability prediction of fault types. It not only reduces the complexity of fault diagnosis, but also improves the efficiency of diagnosis, making the fault diagnosis process of the light source controller more intelligent and systematic, and providing reliable technical support for equipment maintenance and fault warning.

[0048] In one embodiment, dynamic threshold judgment is performed on the probability distribution of fault types to obtain corresponding preliminary fault analysis results, including: The fault type probability data generated by the digital light source controller during operation forms a probability distribution sequence. Based on the probability distribution sequence, the fault types are sorted in descending order of probability value to generate an ordered sequence containing the fault types and their corresponding probability values. This sequence reflects the relative probability of each type of fault.

[0049] Compare and analyze the currently obtained fault type probability sequence with the similar fault probability data stored in the historical database. By calculating the difference between the current probability value and the historical average value, the probability deviation value of each fault type is obtained. These deviation values ​​represent the degree of change in the fault probability relative to the historical level.

[0050] The calculated probability deviation value is divided into multiple levels of threshold intervals according to preset rules. Multiple threshold points are pre-set to divide the deviation value range into high-risk, medium-risk, low-risk and other levels. Each interval corresponds to a different fault risk level, realizing hierarchical management of faults.

[0051] Based on the threshold interval, the occurrence frequency of each fault type in different intervals is statistically analyzed to generate the frequency distribution data of the fault type. This distribution data reflects the distribution characteristics of various faults in different risk level intervals.

[0052] The prediction model is established by using the historical frequency distribution data to predict and analyze the fault development trend. The prediction model uses the time series analysis method to calculate the trend prediction value of fault development in the future based on the historical change law of fault frequency.

[0053] The predicted value is matched with the pre-set fault judgment standard to calculate the matching degree. The judgment standard includes the frequency threshold, change rate and other parameters of various faults. By calculating the degree of match between the predicted value and the standard value, the matching degree index of different fault types is obtained.

[0054] Comprehensively analyze the matching index and frequency distribution characteristics of the fault type, combined with the preset fault diagnosis rules, to form a preliminary fault analysis result. The result includes information such as fault type determination, risk level assessment, and development trend warning, providing a decision-making basis for subsequent fault handling.

[0055] This embodiment achieves accurate identification and early warning of digital light source controller faults by establishing a dynamic threshold judgment mechanism. By sorting and analyzing the probability distribution of fault types and combining historical data comparison, a multi-level fault threshold interval is constructed to improve the accuracy of fault diagnosis. The system predicts trends based on frequency distribution characteristics and establishes a dynamic evaluation mechanism for fault development, making fault warning more forward-looking. Through matching degree calculation and comprehensive analysis, a comprehensive fault diagnosis result is formed, providing a scientific basis for equipment maintenance. It breaks through the limitations of traditional fixed threshold judgment, improves the sensitivity and reliability of fault diagnosis, and effectively reduces the misjudgment rate. At the same time, the adaptive adjustment mechanism of the dynamic threshold enables the system to have stronger environmental adaptability and enhances the practicality of fault diagnosis. This fault diagnosis method based on multi-dimensional analysis not only improves the safety of equipment operation, but also provides strong support for preventive maintenance and extends the service life of equipment.

[0056] In one embodiment, sampling period characteristic fault diagnosis is performed on the preliminary fault analysis result to obtain corresponding fault diagnosis results, including: During the operation of the digital light source controller, sampling period characteristic fault diagnosis is carried out based on the preliminary fault analysis results. The sampling period characteristic fault diagnosis converts the preliminary fault analysis results into a periodic data sequence through period segmentation processing, and the periodic data sequence contains fault characteristic data of multiple sampling periods. During the periodic segmentation processing, the data is divided in units of a fixed time interval T, and the data in each time interval constitutes a sampling period, thereby obtaining a periodic data sequence in the form of {D1, D2, ..., Dn}, where Di represents the data set of the i-th sampling period.

[0057] For the obtained periodic data sequence, fault state matching analysis is performed to identify state transition information. The state matching analysis is based on a preset fault state feature library, matching the data of each sampling cycle with the state pattern in the feature library. The feature library contains multiple state patterns such as normal operation state S0, minor fault state S1, moderate fault state S2, and severe fault state S3. By calculating the similarity between the data features and each state pattern, the fault state corresponding to each sampling cycle is determined, and the state transition relationship between adjacent cycles is recorded.

[0058] The state transition information is processed in segments to form a state interval sequence. The segmentation process is based on state stability, and multiple consecutive identical or similar states are combined into a state interval. The division of the state interval is set with a minimum duration threshold Tmin. Only states whose duration exceeds the minimum duration threshold Tmin will be divided into independent intervals. The state interval sequence reflects the time distribution characteristics of the fault state.

[0059] Based on the state interval sequence, the fault periodicity matching is performed to extract the periodic characteristics of the fault. The periodic matching adopts the sliding window method, and the window size is a specific number of state intervals. By calculating the similarity of the state sequence at different window positions, the recurring state pattern is identified and a vector describing the fault periodic characteristics is generated. The fault period vector contains information such as the period length and the state transition law.

[0060] The fault cycle vector undergoes multiple layers of recursive analysis to obtain a more detailed description of the fault state. Recursive analysis examines fault characteristics at different time scales, including short-term fluctuations, medium-term trends, long-term evolution, and other levels. The number of layers of recursive analysis is determined by the complexity of the fault type. Each layer of analysis will generate a state description sequence, and finally the results of each layer are combined to form a complete fault state sequence.

[0061] The fault state sequence is used to calculate the probability of different types of faults, thereby determining the confidence of fault diagnosis. The probability calculation is based on the Bayesian reasoning method, combined with the prior probability distribution in the historical fault database. For each possible fault type, the corresponding posterior probability is calculated. The higher the probability value, the more reliable the diagnosis result. The calculation of diagnostic confidence also considers factors such as the stability of the state sequence and the significance of the periodic characteristics.

[0062] Finally, the fault diagnosis result is determined based on the fault diagnosis confidence. The diagnosis result contains information such as fault type, severity, and development trend. By setting the confidence threshold, when the confidence of a certain fault type exceeds the confidence threshold, the existence of this type of fault can be confirmed. The diagnosis result also predicts the development trend of the fault, providing a basis for subsequent maintenance decisions.

[0063] This embodiment achieves accurate identification of fault states and dynamic tracking of transfer rules by periodically segmenting the fault characteristics of the digital light source controller, effectively improving the accuracy of fault diagnosis. Based on the multi-layer recursive analysis method, the fault characteristics are deeply mined at different time scales, making the description of the fault state more comprehensive and detailed, providing reliable data support for fault diagnosis. By introducing the Bayesian reasoning mechanism to calculate the probability of fault types and combining the historical fault database, a scientific fault diagnosis confidence assessment system is established, which significantly enhances the credibility of the fault diagnosis results. The state stability judgment and minimum duration threshold control are used to effectively filter out the interference caused by instantaneous fluctuations and ensure the stability of fault diagnosis. This method can not only accurately identify the current fault, but also predict the development trend of the fault, providing an important decision-making basis for equipment maintenance, and has a strong practical value.

[0064] 7. The digital light source controller fault diagnosis method according to claim 1 is characterized in that the fault diagnosis results are graded and analyzed to obtain the corresponding fault handling solution, including: By classifying the faults in the fault diagnosis results, dimensional fault classification data is generated. In terms of the degree of fault impact, it is divided into four levels according to the severity of the impact on system operation: fatal faults, serious faults, general faults and minor faults. Fatal faults cause the system to be completely unable to operate, serious faults cause the loss of major system functions, general faults affect some minor functions, and minor faults have little impact on system operation.

[0065] In terms of fault frequency, faults are divided into high-frequency faults (more than 3 times per month), medium-frequency faults (1-3 times per month) and low-frequency faults (less than 1 time per month) based on the number of faults that occur per unit time.

[0066] In terms of fault duration, faults are divided into long-term faults (more than 24 hours), medium-term faults (4-24 hours) and short-term faults (less than 4 hours) according to the duration of a single fault.

[0067] In terms of the degree of fault hazard, it is divided into major hazard, moderate hazard and minor hazard according to the degree of threat to equipment and personnel safety.

[0068] In terms of repair difficulty, it is divided into high difficulty, medium difficulty and low difficulty according to the technical level and tools and equipment required for repair.

[0069] In terms of fault repair time, it is divided into long-time repair (more than 8 hours), medium-time repair (2-8 hours) and short-time repair (within 2 hours) according to the working hours required for repair.

[0070] A multi-level comprehensive evaluation is performed based on the dimensional fault classification data, and the hierarchical analysis method is used to determine the weight of each dimension. The weight of the fault impact degree is 0.3, the weight of the fault frequency is 0.2, the weight of the fault duration is 0.15, the weight of the fault hazard is 0.15, the weight of the repair difficulty is 0.1, and the weight of the fault repair time is 0.1.

[0071] An evaluation index system is established using a comprehensive evaluation method. The evaluation levels are set as A (extremely serious), B (relatively serious), C (moderate), and D (relatively minor). The comprehensive evaluation value of each fault is obtained through fuzzy matrix calculation to form fault level assessment information.

[0072] The fault handling priority is calculated based on the fault level assessment information. A weighted scoring method is used to incorporate factors such as fault level, system importance, and maintenance resource availability into the scoring system. The scoring rules are set as follows: Fault level A scores 100 points, B scores 80 points, C scores 60 points, and D scores 40 points; the core part of the system scores 100 points, important parts score 80 points, and general parts score 60 points; sufficient maintenance resources score 100 points, resource tension scores 60 points, and resource shortage scores 30 points. The fault handling priority sequence is generated by sorting the total scores.

[0073] Allocate repair resources according to the fault handling priority sequence, and make overall arrangements for personnel, equipment, spare parts and other resources. Formulate standard working hours quotas, and allocate corresponding numbers of maintenance personnel for different levels of faults: 3-4 people for A-level faults, 2-3 people for B-level faults, 1-2 people for C-level faults, and 1 person for D-level faults. Equipment and spare parts are allocated according to the fault type and repair difficulty to form a fault repair resource allocation plan.

[0074] According to the fault repair resource allocation plan, the repair execution sequence is formulated, and the repair process is arranged in sequence using the critical path method. A repair network diagram is drawn, and the earliest start time, the latest start time, and the total time difference of each repair activity are calculated to determine the critical path. Combined with constraints such as maintenance personnel working hours and equipment use plans, a specific fault repair execution sequence is generated.

[0075] For example, a complete fault handling plan is formed by integrating the fault repair resource configuration plan and the fault repair execution sequence. The plan clearly defines the division of labor, equipment deployment, spare parts preparation, repair steps, quality requirements, safety measures, etc., and formulates a detailed progress plan and emergency plan to ensure that the fault repair work is carried out in an orderly manner.

[0076] This embodiment establishes a systematic fault handling mechanism by performing multi-dimensional classification processing and solution analysis on the fault diagnosis results of the digital light source controller. The method adopts a multi-dimensional fault classification data system to conduct a comprehensive assessment from the aspects of fault impact, occurrence frequency, duration, hazard, repair difficulty and repair time, so that the fault classification is more accurate and objective. The multi-level evaluation mechanism based on the hierarchical analysis method and the fuzzy comprehensive evaluation method realizes the scientific quantification of the fault level and provides a reliable basis for fault handling decisions. The fault handling priority is calculated by the weighted scoring method, and the fault level, system importance and resource availability are comprehensively considered to ensure that critical faults are handled in a timely manner. The repair execution sequence is formulated in combination with the critical path method, which realizes the reasonable allocation and efficient utilization of repair resources and significantly improves the fault handling efficiency. At the same time, the integrated fault handling solution clarifies the specific requirements, effectively reduces the uncertainty in the repair process, and ensures the standardization and controllability of fault repair work.

[0077] The present invention further provides a digital light source controller fault diagnosis device, which is applied to any one of the digital light source controller fault diagnosis methods described above, comprising: The acquisition module is used to obtain the original sensor data of the digital light source controller and perform multi-dimensional analysis and extraction to obtain the corresponding multi-dimensional sensor data; An analysis module, which is used to extract time-domain and frequency-domain features from multi-dimensional sensor data to obtain corresponding sensor feature data; The association module is used to perform fault feature analysis on the sensor feature data to obtain the corresponding fault type probability distribution; A processing module, which is used to perform dynamic threshold judgment on the probability distribution of fault types and obtain corresponding preliminary fault analysis results; A control module, the control module is used to perform sampling period characteristic fault diagnosis on the preliminary fault analysis result to obtain a corresponding fault diagnosis result; The execution module is used to classify the fault diagnosis results and perform solution analysis to obtain the corresponding fault handling solution.

[0078] A digital light source controller fault diagnosis device provided by the present invention can comprehensively capture the operating state characteristics of the controller by performing multi-dimensional analysis extraction and feature analysis on the original sensor data of the digital light source controller, significantly improve the accuracy of fault diagnosis, and provide reliable data support for fault prevention and processing. By extracting the time domain and frequency domain features of the sensor data and combining the fault feature analysis, accurate identification and classification of various controller faults are achieved, which helps to quickly locate the cause of the fault and reduce the system downtime. Dynamic threshold judgment based on the probability distribution of the fault type can ensure that the system can timely detect potential faults under different operating conditions, reduce the misjudgment rate, and improve the reliability of the diagnosis system. Through sampling period feature fault diagnosis and hierarchical processing, a more reasonable fault processing plan is formulated, and the classification processing of faults is realized through the hierarchical response strategy, thereby effectively improving the maintenance efficiency of the system and extending the service life of the equipment. At the same time, by performing hierarchical analysis on the fault diagnosis results, the processing strategy can be flexibly adjusted according to the severity and type of different faults, so that the system is more adaptable to complex and changeable application scenarios.

[0079] It should be noted that technicians in the relevant technical field can clearly understand that for the convenience and conciseness of description, the specific working process of the system and each module described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0080] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A digital light source controller fault diagnosis method, characterized in that: include: Acquire the original sensor data of the digital light source controller, and perform multi-dimensional analysis and extraction to obtain corresponding multi-dimensional sensor data; Extracting time-domain and frequency-domain features from the multidimensional sensor data to obtain corresponding sensor feature data; Performing fault feature analysis on the sensor feature data to obtain a corresponding fault type probability distribution; Performing dynamic threshold judgment on the probability distribution of the fault type to obtain corresponding preliminary fault analysis results; Perform sampling period characteristic fault diagnosis on the preliminary fault analysis result to obtain a corresponding fault diagnosis result; The fault diagnosis results are graded and analyzed to obtain corresponding fault handling solutions.

2. The digital light source controller fault diagnosis method according to claim 1, characterized in that: The original sensor data of the digital light source controller is obtained, and multi-dimensional analysis and extraction are performed to obtain corresponding multi-dimensional sensor data, including: Performing data collection on the digital light source controller to obtain original sensor data; Performing frequency band matrix decomposition on the original sensor data to obtain coefficient matrices of different frequency bands; Reconstructing the coefficient matrix to obtain reconstructed signal data; Performing Fourier transform on the reconstructed signal data to obtain frequency spectrum feature data; Performing peak distribution analysis on the frequency spectrum feature data to obtain multi-dimensional sensing features; The multi-dimensional sensing features are subjected to data fusion to obtain the multi-dimensional sensing data.

3. The digital light source controller fault diagnosis method according to claim 1, characterized in that: The extracting time domain and frequency domain features of the multi-dimensional sensor data to obtain corresponding sensor feature data includes: Performing empirical mode decomposition on the multidimensional sensing data to obtain a light source wave mode function sequence; Performing time domain feature calculation according to the light source wave mode function sequence to obtain a light source time-varying feature sequence; Performing Hilbert transform on the light source wave mode function sequence to obtain the light source instantaneous spectrum; Performing local linear embedding matrix construction on the instantaneous spectrum of the light source to obtain a light source state characteristic matrix; Performing bispectral analysis according to the instantaneous spectrum of the source to obtain a light source fault vector; The light source state feature matrix and the light source fault vector are sparsely autoencoded and fused to obtain the sensing feature data.

4. The digital light source controller fault diagnosis method according to claim 1, characterized in that: The performing fault feature analysis on the sensor feature data to obtain a corresponding fault type probability distribution includes: The sensing characteristic data is subjected to wavelet packet decomposition processing to obtain a multi-scale decomposition coefficient of a light source fault; Calculate the energy entropy eigenvalue according to the multi-scale decomposition coefficient of the light source fault to obtain the light source fault energy eigenvector; Performing adaptive cluster analysis on the light source fault energy feature vector to obtain a light source control feature cluster center; Constructing a Mahalanobis distance matrix according to the light source control feature clustering center to obtain light source controller feature distance data; Performing hierarchical analysis on the characteristic distance data of the light source controller to obtain the correlation degree of the light source fault characteristics; Perform fault mode identification according to the light source fault feature correlation to obtain a light source controller fault identification result; Establishing a radial basis function according to the light source controller fault identification result to obtain a light source controller fault probability density function; The fault type probability is calculated for the light source fault probability density function to obtain the fault type probability distribution.

5. The digital light source controller fault diagnosis method according to claim 1, characterized in that: The dynamic threshold judgment is performed on the probability distribution of the fault type to obtain the corresponding preliminary fault analysis result, including: Sorting the probability of fault types according to the fault type probability distribution to obtain a fault type probability sequence; Performing historical data comparison on the fault type probability sequence to obtain a fault type probability deviation value; Performing multi-level threshold division on the fault type probability deviation value to obtain a fault type threshold interval; Performing classification statistics of fault types according to the fault type threshold interval to obtain a fault type frequency distribution; Performing trend prediction on the frequency distribution of the fault types to obtain a predicted value of the fault development trend; Calculating the matching degree of the fault development trend prediction value according to the preset fault judgment standard to obtain the fault type matching degree; Fault analysis is performed on the fault type matching degree and the fault type frequency distribution to obtain the preliminary fault analysis result.

6. The digital light source controller fault diagnosis method according to claim 1, characterized in that: The performing sampling period characteristic fault diagnosis on the preliminary fault analysis result to obtain a corresponding fault diagnosis result includes: Performing period segmentation on the preliminary fault analysis result to obtain a corresponding periodic data sequence; Perform fault state matching analysis according to the periodic data sequence to obtain corresponding state transition information; Processing the state transition information in segments to obtain a corresponding state interval sequence; Perform fault periodicity matching according to the state interval sequence to obtain a corresponding fault period vector; Performing multi-layer recursive analysis on the fault cycle vector to obtain a corresponding fault state sequence; Performing type probability calculation according to the fault state sequence to obtain corresponding fault diagnosis confidence; Fault diagnosis is performed on the fault diagnosis confidence to obtain the fault diagnosis result.

7. The digital light source controller fault diagnosis method according to claim 1, characterized in that: The fault diagnosis results are graded and analyzed to obtain corresponding fault handling solutions, including: Performing fault classification on the fault diagnosis results to obtain fault classification data including fault impact degree, fault frequency, fault duration, fault severity, repair difficulty and fault repair time; Perform a multi-level comprehensive evaluation based on the dimensional fault classification data to obtain fault level evaluation information; Performing fault priority sorting calculation on the fault level assessment information to obtain a fault processing priority sequence; Allocate repair resources according to the fault handling priority sequence to obtain a fault repair resource configuration plan; Performing repair arrangement analysis according to the fault repair resource configuration plan to obtain a fault repair execution sequence; The fault repair resource configuration scheme and the fault repair execution sequence are integrated to obtain the fault handling scheme.

8. A digital light source controller fault diagnosis device, characterized in that: The digital light source controller fault diagnosis method applied to any one of claims 1 to 7 above comprises: An acquisition module, which is used to acquire original sensor data of the digital light source controller and perform multi-dimensional analysis and extraction to obtain corresponding multi-dimensional sensor data; An analysis module, the analysis module is used to extract time domain and frequency domain features of the multi-dimensional sensor data to obtain corresponding sensor feature data; An association module, the association module is used to perform fault feature analysis on the sensor feature data to obtain a corresponding fault type probability distribution; A processing module, the processing module is used to perform dynamic threshold judgment on the probability distribution of the fault type to obtain a corresponding preliminary fault analysis result; A control module, the control module is used to perform sampling period characteristic fault diagnosis on the preliminary fault analysis result to obtain a corresponding fault diagnosis result; An execution module is used to perform hierarchical processing on the fault diagnosis results and perform solution analysis to obtain a corresponding fault processing solution.

Images