Fault diagnosis system and method for mining frequency conversion speed regulation system

Abstract

The invention relates to the technical field of electromagnetic variables, and discloses a fault diagnosis system and method for a mining frequency conversion speed regulation system, and the system comprises an information collection module, a matrix construction module, a high-dimensional vector module, a low-dimensional feature module, a fault diagnosis module and a diagnosis report module, and is used for collecting an original multi-source heterogeneous signal set; filtering the noise, and performing dimension reconstruction to obtain a standardized signal matrix; extracting a high-order statistical moment, evaluating an energy distribution feature, and analyzing an energy entropy to construct a holographic feature vector; mapping the holographic feature vectors to a preset feature importance evaluation network, and rearranging the holographic feature vectors to obtain a low-dimensional feature subset; carrying out confidence evaluation on the low-dimensional feature subset, and carrying out decision fusion to obtain preliminary fault diagnosis; performing time sequence comparison on the preliminary fault diagnosis and a historical health state baseline to obtain a final fault type, severity and possible positioning information so as to generate a structured diagnosis report; according to the invention, the fault diagnosis efficiency of the mining frequency conversion speed regulation system can be improved.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic variable technology, and in particular to a fault diagnosis system and method for mine variable frequency speed control systems. Background Technology

[0002] In the fault diagnosis process of mining variable frequency speed control systems, existing technologies lack precise time base synchronization design for the acquisition of multi-source heterogeneous signals such as three-phase output current, DC bus voltage, and power device temperature. The sampling rhythm of different types of signals is inconsistent, resulting in timing misalignment of the acquired raw data. This makes it impossible to accurately reproduce the correlation between various parameters under the system's operating state, thus creating potential deviations for subsequent fault feature extraction.

[0003] Existing technologies fail to implement differentiated noise filtering for the rapid and slow-changing characteristics of signals in mine variable frequency speed control systems, making it difficult to completely eliminate interference from high-frequency background noise and low-frequency drift noise. At the same time, the lack of effective importance assessment and redundancy removal mechanisms for extracted features results in an excessively high proportion of invalid information in the feature vectors. This not only increases the processing load of fault diagnosis but also directly affects the accuracy and timeliness of fault judgment. Therefore, how to improve the efficiency of fault diagnosis in mine variable frequency speed control systems has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a fault diagnosis system and method for mine variable frequency speed control systems to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a fault diagnosis system for a mining variable frequency speed control system, characterized in that the system includes an information acquisition module, a matrix construction module, a high-dimensional vector module, a low-dimensional feature module, a fault diagnosis module, and a diagnosis report module, wherein:

[0006] The information acquisition module is used to synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system, so as to integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system.

[0007] The matrix construction module is used to filter noise of different dimensions in the original multi-source heterogeneous signal set, and to reconstruct the dimensions of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system.

[0008] The high-dimensional vector module is used to extract the higher-order statistical moments of the time-domain waveform in the standardized signal matrix, evaluate the energy distribution characteristics of the spectral changes in the standardized signal matrix, and analyze the energy entropy of the time-frequency components in the standardized signal matrix to construct the holographic feature vector of the mining variable frequency speed control system.

[0009] The low-dimensional feature module is used to map the holographic feature vector to a preset feature importance evaluation network, and to rearrange the features in the holographic feature vector whose feature dimension weight is lower than the standard threshold in the preset feature importance evaluation network to obtain a low-dimensional feature subset of the mining variable frequency speed control system.

[0010] The fault diagnosis module is used to evaluate the confidence level of the low-dimensional feature subset and perform decision fusion on the evaluation results to obtain the preliminary fault diagnosis of the mining variable frequency speed control system.

[0011] The diagnostic report module is used to compare the preliminary fault diagnosis with the historical health status baseline of the mining variable frequency speed control system in a time series to obtain the final fault type, severity and possible location information of the mining variable frequency speed control system, so as to generate a structured diagnostic report of the mining variable frequency speed control system.

[0012] In a preferred embodiment, when the information acquisition module synchronously acquires the three-phase output current, DC bus voltage, and power device temperature signals of the mining variable frequency speed control system to integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system, it is specifically used for:

[0013] Synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system to obtain the multi-source signals of the mining variable frequency speed control system;

[0014] The DC bus voltage is used as the time base synchronization reference signal for the multi-source signal, and the target sampling interval of the mining variable frequency speed control system is determined based on the period of the time base synchronization reference signal.

[0015] Based on the target sampling interval, the original discrete sequence of the three-phase output current is resampled using time-base synchronization to obtain the synchronized three-phase output current of the mining variable frequency speed control system.

[0016] The synchronized three-phase output current, the time base synchronization reference signal, and the power device temperature signal are time-aligned to obtain the original multi-source heterogeneous signal set of the mining variable frequency speed control system.

[0017] In a preferred embodiment, when the matrix construction module performs filtering of noise of different dimensions in the original multi-source heterogeneous signal set and dimensional reconstruction of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system, it is specifically used for:

[0018] Based on the signal change rate of the mining variable frequency speed control system, the signals in the original multi-source heterogeneous signal set are divided into fast-changing dimension signals and slow-changing dimension signals.

[0019] Based on the amplitude change between adjacent sampling points in the rapidly changing dimension signal, a first dynamic filtering threshold for the rapidly changing dimension signal is set.

[0020] Signal points in the rapidly changing dimensional signal whose amplitude changes are lower than the first dynamic filtering threshold are treated as high-frequency background noise and filtered out to obtain the first filtered signal of the original multi-source heterogeneous signal set.

[0021] Within a preset time window, a second dynamic filtering threshold for the slowly varying dimension signal is set based on the fluctuation range of the slowly varying dimension signal.

[0022] Signal segments whose fluctuation range is consistently below the second dynamic filtering threshold are treated as low-frequency drift noise and filtered out to obtain the second filtered signal of the slowly varying dimension signal.

[0023] In a preferred embodiment, when the matrix construction module performs dimensional reconstruction on the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system, it is specifically used for:

[0024] The first filtered signal and the second filtered signal are bidirectionally coupled to obtain the standard filtered signal of the mining variable frequency speed control system.

[0025] The standard filtered signal is mapped to the multi-dimensional data space of the mining variable frequency speed control system;

[0026] Within the multidimensional data space, the aligned signal segment sequence of the standard filtered signal is obtained based on the relative proportions of the alignment and amplitude range of the signal segments on the time axis.

[0027] The amplitude normalization process is performed on the aligned signal segment sequence to obtain the normalized signal sequence of the mining variable frequency speed control system;

[0028] The normalized signal sequence is reconstructed to obtain the standardized signal matrix of the mining variable frequency speed control system.

[0029] In a preferred embodiment, the high-dimensional vector module, when performing the extraction of high-order statistical moments of the time-domain waveform from the standardized signal matrix, is specifically used for:

[0030] Identify the periodic zero-crossing points of the standardized time-domain current waveform in the standardized signal matrix;

[0031] Using the periodic zero-crossing point as the dividing criterion, the standardized time-domain current waveform is divided into current signal period segments of the standardized time-domain current waveform;

[0032] By aggregating the DC components of the current sampling points in the current signal periodic segment, the DC bias of the current signal periodic segment is obtained;

[0033] Based on the DC bias, the current sampling point is biased and corrected to obtain the corrected period current signal of the standardized time-domain current waveform.

[0034] The asymmetry of amplitude distribution in the current signal of the correction period is analyzed to obtain the waveform distortion characteristics of the amplitude distribution, which are used as the higher-order statistical moments of the standardized signal matrix.

[0035] In a preferred embodiment, the high-dimensional vector module, when performing the evaluation of the energy distribution characteristics of the spectral changes in the normalized signal matrix, is specifically used for:

[0036] The current signal of the correction period segment is subjected to spectral transformation to obtain the period segment spectrum of the current signal of the correction period segment;

[0037] The fundamental energy of the frequency band is determined based on the sum of the squares of the amplitudes of the spectral components in the frequency band of the periodic segment spectrum.

[0038] The basic energy is integrated into the total energy of the periodic spectrum;

[0039] The distribution ratio vector of the overall energy is nonlinearly fused to obtain the energy distribution characteristics of the standardized signal matrix, wherein the calculation formula of the energy distribution characteristics is as follows:

[0040] ;

[0041] In the formula, The energy distribution characteristics are as described above. It is an exponential function. The total number of the frequency bands. For the first The fundamental energy of each frequency band, The preset stability coefficient, It is a logarithmic function.

[0042] In a preferred embodiment, the high-dimensional vector module, when performing the analysis of the energy entropy of the time-frequency components in the normalized signal matrix, is specifically used for:

[0043] The current signal of the correction period segment is subjected to joint time-frequency reconstruction to obtain the time-frequency distribution matrix of the current signal of the correction period segment;

[0044] On the time-frequency plane, the time-frequency distribution matrix is ​​divided into a rectangular analysis region for the mining variable frequency speed control system;

[0045] The total energy of the rectangular analysis area is obtained by summing the energy values ​​of all time-frequency components within the rectangular analysis area.

[0046] Based on the total energy of the region, the energy entropy of the standardized signal matrix is ​​analyzed, wherein the formula for calculating the energy entropy is as follows:

[0047] ;

[0048] In the formula, The energy entropy, This represents the total number of rectangular analysis regions. For the first The proportion of energy in each rectangular analysis region to the total energy of the region. It is a logarithmic function.

[0049] In a preferred embodiment, when the low-dimensional feature module performs the operation of mapping the holographic feature vector to a preset feature importance evaluation network and rearranging the features whose feature dimension weights in the holographic feature vector are lower than the standard threshold in the preset feature importance evaluation network to obtain the low-dimensional feature subset of the mining variable frequency speed control system, it is specifically used for:

[0050] Based on a preset feature importance evaluation network, the importance of the holographic feature vector is evaluated to obtain the initial weight values ​​of the feature dimensions in the holographic feature vector;

[0051] The initial weight values ​​of the feature dimensions are dynamically adjusted to obtain the adjusted weight values ​​of the feature dimensions.

[0052] Based on the overall distribution characteristics of the corrected weight values, the weight screening threshold of the preset feature importance evaluation network is determined;

[0053] The corrected weight value is mapped to the weight screening threshold to filter out low-dimensional features whose feature dimension is lower than the weight screening threshold, thereby obtaining the feature dimension to be reconstructed in the holographic feature vector;

[0054] Analyze the intrinsic correlation between the feature dimension to be reconstructed and the high-weight feature dimension whose corrected weight value is not lower than the weight screening threshold;

[0055] Based on the inherent correlation, the feature dimension to be reconstructed is directionally fused to the high-weight feature dimension to obtain a low-dimensional feature subset of the mining variable frequency speed control system.

[0056] In a preferred embodiment, when the fault diagnosis module performs confidence assessment on the low-dimensional feature subset and performs decision fusion on the assessment results to obtain a preliminary fault diagnosis of the mining variable frequency speed control system, it is specifically used for:

[0057] The matching degree analysis is performed between the simplified features in the low-dimensional feature subset and the indication strength of the preset fault type to obtain the first layer confidence of the low-dimensional feature subset;

[0058] The first layer confidence score is subjected to time smoothing correction to obtain the second layer confidence score of the low-dimensional feature subset;

[0059] The first layer confidence and the second layer confidence are jointly corrected to obtain the target feature confidence of the low-dimensional feature subset;

[0060] Based on the confidence level of the target features, determine the comprehensive probability of state-related faults of the mining variable frequency speed control system;

[0061] The overall probability of the state fault is mapped to the preset fault type to obtain the preliminary fault diagnosis of the mining variable frequency speed control system.

[0062] To address the above problems, this invention also provides a fault diagnosis method for a mining variable frequency speed control system, the method comprising:

[0063] S1. Synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system, and integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system;

[0064] S2. Filter the noise in different dimensions of the original multi-source heterogeneous signal set, and reconstruct the dimensions of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system;

[0065] S3. Extract the higher-order statistical moments of the time-domain waveform in the standardized signal matrix, evaluate the energy distribution characteristics of the spectral changes in the standardized signal matrix, and analyze the energy entropy of the time-frequency components in the standardized signal matrix to construct the holographic feature vector of the mining variable frequency speed control system;

[0066] S4. Map the holographic feature vector to a preset feature importance evaluation network, and rearrange the features in the holographic feature vector whose feature dimension weight is lower than the standard threshold in the preset feature importance evaluation network to obtain a low-dimensional feature subset of the mining variable frequency speed control system;

[0067] S5. Confidence assessment is performed on the low-dimensional feature subset, and decision fusion is performed on the assessment results to obtain the preliminary fault diagnosis of the mining variable frequency speed control system.

[0068] S6. The preliminary fault diagnosis is compared with the historical health status baseline of the mining variable frequency speed control system in a time series to obtain the final fault type, severity and possible location information of the mining variable frequency speed control system, so as to generate a structured diagnostic report of the mining variable frequency speed control system.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] 1. This invention synchronously acquires the three-phase output current, DC bus voltage, and power device temperature signals of a mining variable frequency speed control system. Using the DC bus voltage as a reference, it achieves time-base synchronized resampling and timing alignment, ensuring the consistency of the original multi-source heterogeneous signal set. Differential noise filtering is implemented for signals with different rates of change, and a standardized signal matrix is ​​formed by combining bidirectional coupling and dimensional reconstruction, maximizing the retention of effective signal information and improving the accuracy of subsequent feature extraction.

[0071] 2. This invention constructs a holographic feature vector by extracting higher-order statistical moments, evaluating energy distribution characteristics, and analyzing energy entropy. A low-dimensional feature subset is then obtained through feature importance assessment and targeted fusion, reducing interference from invalid information. Through confidence assessment, decision fusion, and time-series comparison with historical health status baselines, the fault type, severity, and possible location are accurately determined, generating a structured diagnostic report and comprehensively improving the timeliness and reliability of fault diagnosis. Attached Figure Description

[0072] Figure 1 This is a system architecture diagram of a fault diagnosis system for a mining variable frequency speed control system provided in an embodiment of the present invention;

[0073] Figure 2 This is a flowchart illustrating a fault diagnosis method for a mining variable frequency speed control system according to an embodiment of the present invention.

[0074] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0077] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0078] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0079] In practice, the server-side equipment deployed in the fault diagnosis system for mine variable frequency speed control systems may consist of one or more devices. This fault diagnosis system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node, providing fault diagnosis services to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage each user terminal. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide fault diagnosis services to various user terminals.

[0080] In terms of implementation, the fault diagnosis system for mine variable frequency speed control systems and the user terminal are mutually compatible. That is, if the fault diagnosis system for mine variable frequency speed control systems is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the fault diagnosis system for mine variable frequency speed control systems is implemented as a website, then the user terminal is implemented as a webpage; or if the fault diagnosis system for mine variable frequency speed control systems is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0081] like Figure 1 The figure shown is a system architecture diagram of a fault diagnosis system for a mining variable frequency speed control system provided in an embodiment of the present invention.

[0082] The fault diagnosis system 100 for a mining variable frequency speed control system described in this invention can be located on a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed on the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed as a website. Depending on the functions implemented, the fault diagnosis system 100 for a mining variable frequency speed control system may include an information acquisition module 101, a matrix construction module 102, a high-dimensional vector module 103, a low-dimensional feature module 104, a fault diagnosis module 105, and a diagnosis report module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device's processor and perform a fixed function, stored in the electronic device's memory.

[0083] In this embodiment of the invention, in the fault diagnosis system for a mining variable frequency speed control system, each of the above-mentioned modules can be implemented independently and can call other modules. Here, "calling" can be understood as a module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the fault diagnosis system for a mining variable frequency speed control system provided by this embodiment of the invention, the applicable scope of the system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the fault diagnosis system. In practical applications, the above-mentioned modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0084] The following describes, with reference to specific embodiments, each component of the fault diagnosis system for mine variable frequency speed control systems and its specific workflow:

[0085] The information acquisition module 101 is used to synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system, so as to integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system.

[0086] In this embodiment of the invention, when the information acquisition module synchronously acquires the three-phase output current, DC bus voltage, and power device temperature signals of the mining variable frequency speed control system to integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system, it is specifically used for:

[0087] Synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system to obtain the multi-source signals of the mining variable frequency speed control system;

[0088] The DC bus voltage is used as the time base synchronization reference signal for the multi-source signal, and the target sampling interval of the mining variable frequency speed control system is determined based on the period of the time base synchronization reference signal.

[0089] Based on the target sampling interval, the original discrete sequence of the three-phase output current is resampled using time-base synchronization to obtain the synchronized three-phase output current of the mining variable frequency speed control system.

[0090] The DC bus voltage of the mining variable frequency speed control system is obtained by rectifying the power frequency AC power. The period of its voltage ripple is directly related to the power frequency period. For example, if the frequency standard is 50Hz, the corresponding power frequency period is 20ms. Therefore, the bus voltage ripple period of this system is usually equal to the power frequency period or an integer fraction of the power frequency period. It can be determined that the two are strongly correlated.

[0091] The synchronized three-phase output current, the time base synchronization reference signal, and the power device temperature signal are time-aligned to obtain the original multi-source heterogeneous signal set of the mining variable frequency speed control system.

[0092] By deploying dedicated signal acquisition elements on the three-phase output terminals, DC bus interface, and power device surfaces of the mining variable frequency speed control system, and simultaneously initiating signal capture operations, the system continuously tracks the real-time waveform changes of the three-phase output current, the fluctuation state of the DC bus voltage, and the continuous temperature values ​​of the power devices. This ensures that no three types of signals are missed or delayed during the acquisition process, and all real-time acquired signal data are completely summarized to obtain the multi-source signals of the mining variable frequency speed control system.

[0093] Differentiate the signal type and set the sampling rate: Switching frequency, harmonic-related current and voltage signals are fast-changing signals and must follow the Nyquist sampling theorem. The sampling frequency should be at least twice the switching frequency, usually 5-10 times, to retain the complete information of the switching frequency and its harmonics; power device temperature and other signals are slowly changing signals and can use a lower sampling rate to avoid generating redundant data.

[0094] Dynamic filtering preserves effective features: For rapidly changing dimension signals, a first dynamic filtering threshold is set to filter out high-frequency background noise and retain effective abrupt changes in switching frequency and harmonics; for slowly changing dimension signals, a second dynamic filtering threshold is set to filter out low-frequency drift noise, ensuring the stability of signals such as temperature and preventing effective information from being covered by noise.

[0095] Time-base synchronization resampling: The target sampling interval is determined based on the DC bus voltage period. The original discrete sequence of the three-phase output current is resampled using time-base synchronization to ensure the timing alignment of multi-source signals and avoid the loss of harmonic and switching frequency related features due to timing misalignment.

[0096] The target sampling interval is determined as follows: using the DC bus voltage as the synchronization reference signal, its period (e.g., a 20ms power frequency period) is determined. Combining the highest frequency components of the fast-changing signal (switching frequency, harmonic frequency), the minimum sampling frequency that satisfies the Nyquist sampling theorem is calculated. The target sampling interval is the reciprocal of the minimum sampling frequency. At the same time, the timing alignment requirements of multi-source signals are also taken into account. Finally, a unified sampling interval is determined. In engineering applications, if the switching frequency is several kHz to tens of kHz, the sampling frequency is usually set to 5-10 times the switching frequency, and the corresponding target sampling interval can be as low as microseconds.

[0097] The DC bus voltage signal is selected as the time base synchronization reference signal from the aggregated multi-source signals. The waveform fluctuations of the signal are continuously tracked, and the complete change process from peak to peak and from valley to valley is recorded. The stable repetition time of the signal is determined as the period. Based on this period, and combined with the transmission characteristics and acquisition requirements of the three types of signals, a unified time interval is set that can keep the acquisition rhythm of all signals consistent. This unified interval is the target sampling interval of the mine variable frequency speed control system.

[0098] Using the predetermined target sampling interval as a fixed standard, the previously acquired original discrete sequence of three-phase output current is systematically reviewed point by point. The timestamp corresponding to each sampling point is checked one by one and accurately compared with the time node of the target sampling interval. Sampling points whose timestamps perfectly match the target sampling interval are retained. For sampling points missing within the target sampling interval, the corresponding data is calculated and supplemented by linear interpolation based on the amplitude change trend of adjacent valid sampling points. Excessive sampling points that exceed the target sampling interval are directly removed. Finally, the synchronized three-phase output current of the mining variable frequency speed control system is obtained.

[0099] Using the time axis of the time base synchronization reference signal as a unified reference, the timestamp information corresponding to each data point in the synchronized three-phase output current, the time base synchronization reference signal, and the power device temperature signal is extracted respectively. The timestamps of the three types of signals are sorted according to the same time scale, and the signal data at the same time node are matched one by one. The time axis is shifted and adjusted for signal data with time offset to ensure that the three types of signals have corresponding accurate data at the same time node, so as to achieve complete and accurate matching of the three types of signals in the time dimension. Finally, the matched three types of signals are integrated to form the original multi-source heterogeneous signal set of the mining variable frequency speed control system.

[0100] The beneficial effects are that it synchronously acquires three key signals of the mining variable frequency speed control system, uses the DC bus voltage as the time base synchronization reference signal and determines the target sampling interval accordingly, and ensures the accurate matching of the synchronized three-phase output current, the time base synchronization reference signal and the power device temperature signal in the time dimension through time base synchronization resampling and timing alignment operations. This effectively avoids signal timing misalignment and data deviation, and enables the integrated original multi-source heterogeneous signal set to completely retain the real state information during system operation. This provides high-quality and highly consistent data support for subsequent matrix construction, feature extraction and fault diagnosis, ensuring the accuracy and reliability of the overall fault diagnosis process.

[0101] The matrix construction module 102 is used to filter noise of different dimensions in the original multi-source heterogeneous signal set and to reconstruct the dimensions of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system.

[0102] In this embodiment of the invention, when the matrix construction module performs filtering of noise of different dimensions in the original multi-source heterogeneous signal set and dimensional reconstruction of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system, it is specifically used for:

[0103] Based on the signal change rate of the mining variable frequency speed control system, the signals in the original multi-source heterogeneous signal set are divided into fast-changing dimension signals and slow-changing dimension signals.

[0104] Based on the amplitude change between adjacent sampling points in the rapidly changing dimension signal, a first dynamic filtering threshold for the rapidly changing dimension signal is set.

[0105] Signal points in the rapidly changing dimensional signal whose amplitude changes are lower than the first dynamic filtering threshold are treated as high-frequency background noise and filtered out to obtain the first filtered signal of the original multi-source heterogeneous signal set.

[0106] Within a preset time window, a second dynamic filtering threshold for the slowly varying dimension signal is set based on the fluctuation range of the slowly varying dimension signal.

[0107] Signal segments whose fluctuation range is consistently below the second dynamic filtering threshold are treated as low-frequency drift noise and filtered out to obtain the second filtered signal of the slowly varying dimension signal.

[0108] When the matrix construction module performs dimensional reconstruction on the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system, it is specifically used for:

[0109] The first filtered signal and the second filtered signal are bidirectionally coupled to obtain the standard filtered signal of the mining variable frequency speed control system.

[0110] The standard filtered signal is mapped to the multi-dimensional data space of the mining variable frequency speed control system;

[0111] Within the multidimensional data space, the aligned signal segment sequence of the standard filtered signal is obtained based on the relative proportions of the alignment and amplitude range of the signal segments on the time axis.

[0112] The amplitude normalization process is performed on the aligned signal segment sequence to obtain the normalized signal sequence of the mining variable frequency speed control system;

[0113] The normalized signal sequence is reconstructed to obtain the standardized signal matrix of the mining variable frequency speed control system.

[0114] By continuously monitoring the amplitude changes of each signal in the original multi-source heterogeneous signal set, and comparing the frequency of change and amplitude fluctuation of different signals within the same time span, we can clearly distinguish between signals that change rapidly and have frequent amplitude fluctuations and signals that change slowly and have mild amplitude fluctuations. The former are classified as fast-changing dimension signals and the latter are classified as slow-changing dimension signals.

[0115] The amplitude data of two adjacent sampling points in the rapidly changing dimension signal are extracted one by one. The amplitude difference between each pair of adjacent sampling points is calculated. The overall distribution characteristics of the amplitude change amplitude of all adjacent sampling point pairs are statistically analyzed. Combined with the normal operating range of the rapidly changing dimension signal, a value that can distinguish between effective signal change and high-frequency noise is selected and set as the first dynamic filtering threshold of the rapidly changing dimension signal.

[0116] The amplitude change of each sampling point in the rapidly changing dimension signal is checked point by point. Sampling points with a change less than the first dynamic filtering threshold are identified as high-frequency background noise. The rapidly changing dimension signal is purified by directly removing these noise sampling points. Sampling points with amplitude changes that conform to the characteristics of effective signals are retained. After integration, the first filtered signal of the original multi-source heterogeneous signal set is obtained.

[0117] A continuous and fixed-length signal segment is selected as a preset time window. All amplitude data of the slowly varying dimension signal are extracted within each time window. The maximum and minimum amplitude values ​​within the window are found, and the difference between them is calculated to determine the fluctuation range of the slowly varying dimension signal within the window. Based on the statistical results of the fluctuation range of multiple consecutive time windows, a value that can distinguish between effective signal fluctuations and low-frequency drift noise is set as the second dynamic filtering threshold for the slowly varying dimension signal.

[0118] The fluctuation of the slowly varying dimension signal is continuously tracked, and the fluctuation range of the signal segment is checked segment by segment. When the fluctuation range of a certain signal segment is always lower than the second dynamic filtering threshold in multiple consecutive time units, the signal segment is determined to be low-frequency drift noise. The slowly varying dimension signal is processed by truncating and removing the noise signal segment, and the part whose fluctuation range meets the characteristics of the effective signal is retained. After integration, the second filtered signal of the slowly varying dimension signal is obtained.

[0119] Correlation analysis is performed on the first and second filtered signals. A one-to-one correspondence is established based on their timestamp information. The two types of signal data at the same time node are mutually verified, and the missing valid information of one side is supplemented. Inconsistent abnormal data is eliminated. Through this bidirectional complementary integration method, a standard filtered signal for the mining variable frequency speed control system with the core characteristics of both types of signals is formed.

[0120] A multi-dimensional data space for a mining variable frequency speed control system is constructed with signal type, time dimension, and amplitude as the core dimensions. Each data point in the standard filtered signal is mapped to the corresponding coordinate position in the multi-dimensional data space according to its signal type, corresponding time marker, and specific amplitude value. This enables the standard filtered signal to form a structured distribution in space, thus completing the mapping from the standard filtered signal to the multi-dimensional data space of the mining variable frequency speed control system.

[0121] Within the multidimensional data space, signal segments of the standard filtered signal are extracted one by one. The start and end points of each signal segment are compared, and signal segments with time deviations are adjusted to ensure that the start and end times of all signal segments on the time axis are consistent. At the same time, the maximum and minimum amplitude values ​​of each signal segment are calculated to ensure that the relative proportions of the amplitude ranges of different signal segments conform to the normal operating rules of the system. The processed signal segments are integrated to obtain the aligned signal segment sequence of the standard filtered signal.

[0122] Traverse all data points in the aligned signal segment sequence, find the maximum and minimum amplitude values ​​in the sequence, calculate the difference between the maximum and minimum amplitude values ​​to obtain the total amplitude range, subtract the minimum amplitude value from the amplitude value of each data point, and then divide by the total amplitude range to adjust the amplitude values ​​of all data points to a fixed range while keeping the relative amplitude relationship between each data point unchanged, and finally obtain the normalized signal sequence of the mining variable frequency speed control system.

[0123] The normalized signal sequence is rearranged according to time order and signal type. Different types of signal data at the same time point are grouped into a row, and different time points are grouped into a column. All normalized signal data are arranged in sequence to form a structured form with clear rows and columns. Each element clearly corresponds to a specific time point and signal type. Through this orderly organization, the dimensional reconstruction of the normalized signal sequence is completed, and the standardized signal matrix of the mining variable frequency speed control system is obtained.

[0124] The beneficial effects are as follows: based on the rate of change of the signal in the mine variable frequency speed control system, the signal is divided into fast-changing dimension signal and slow-changing dimension signal. By setting targeted dynamic filtering thresholds, high-frequency background noise and low-frequency drift noise are filtered out respectively, effectively purifying the first and second filtered signals and preserving the effective information in the signal to the greatest extent. Then, a standard filtered signal is obtained through bidirectional coupling. Through operations such as mapping multi-dimensional data space, aligning signal segments, amplitude normalization, and dimension reconstruction, the signal maintains a high degree of consistency and structured characteristics in the time axis and amplitude range. The resulting standardized signal matrix data has high quality and strong correlation, providing accurate and reliable basic data support for subsequent steps such as high-order statistical moment extraction, energy distribution feature evaluation, and holographic feature vector construction, ensuring the stability and accuracy of the overall fault diagnosis process.

[0125] The high-dimensional vector module 103 is used to extract the high-order statistical moments of the time-domain waveform in the standardized signal matrix, evaluate the energy distribution characteristics of the spectral changes in the standardized signal matrix, and analyze the energy entropy of the time-frequency components in the standardized signal matrix to construct the holographic feature vector of the mining variable frequency speed control system.

[0126] In this embodiment of the invention, when the high-dimensional vector module extracts the higher-order statistical moments of the time-domain waveform from the standardized signal matrix, it is specifically used for:

[0127] Identify the periodic zero-crossing points of the standardized time-domain current waveform in the standardized signal matrix;

[0128] Using the periodic zero-crossing point as the dividing criterion, the standardized time-domain current waveform is divided into current signal period segments of the standardized time-domain current waveform;

[0129] By aggregating the DC components of the current sampling points in the current signal periodic segment, the DC bias of the current signal periodic segment is obtained;

[0130] Based on the DC bias, the current sampling point is biased and corrected to obtain the corrected period current signal of the standardized time-domain current waveform.

[0131] The asymmetry of amplitude distribution in the current signal of the correction period is analyzed to obtain the waveform distortion characteristics of the amplitude distribution, which are used as the higher-order statistical moments of the standardized signal matrix.

[0132] The high-dimensional vector module, when evaluating the energy distribution characteristics of spectral variations in the normalized signal matrix, is specifically used for:

[0133] The current signal of the correction period segment is subjected to spectral transformation to obtain the period segment spectrum of the current signal of the correction period segment;

[0134] The fundamental energy of the frequency band is determined based on the sum of the squares of the amplitudes of the spectral components in the frequency band of the periodic segment spectrum.

[0135] The basic energy is integrated into the total energy of the periodic spectrum;

[0136] The distribution ratio vector of the overall energy is nonlinearly fused to obtain the energy distribution characteristics of the standardized signal matrix, wherein the calculation formula of the energy distribution characteristics is as follows:

[0137] ;

[0138] In the formula, The energy distribution characteristics are as described above. It is an exponential function. The total number of the frequency bands. For the first The fundamental energy of each frequency band, The preset stability coefficient, It is a logarithmic function.

[0139] When performing the analysis of the energy entropy of the time-frequency components in the normalized signal matrix, the high-dimensional vector module is specifically used for:

[0140] The current signal of the correction period segment is subjected to joint time-frequency reconstruction to obtain the time-frequency distribution matrix of the current signal of the correction period segment;

[0141] On the time-frequency plane, the time-frequency distribution matrix is ​​divided into a rectangular analysis region for the mining variable frequency speed control system;

[0142] The total energy of the rectangular analysis area is obtained by summing the energy values ​​of all time-frequency components within the rectangular analysis area.

[0143] Based on the total energy of the region, the energy entropy of the standardized signal matrix is ​​analyzed, wherein the formula for calculating the energy entropy is as follows:

[0144] ;

[0145] In the formula, The energy entropy, This represents the total number of rectangular analysis regions. For the first The proportion of energy in each rectangular analysis region to the total energy of the region. It is a logarithmic function.

[0146] The normalized time-domain current waveform in the normalized signal matrix is ​​continuously tracked, and the amplitude change of the waveform is checked point by point. The signal points where the amplitude changes from positive to negative or from negative to positive and just passes through zero are identified. At the same time, the occurrence time of these signal points is recorded, and the time interval between adjacent signal points of the same type is verified to remain stable and consistent. The zero-value crossing points that conform to the stable interval law are determined as the periodic zero-crossing points of the normalized time-domain current waveform.

[0147] Using the identified periodic zero-crossing points as the dividing boundaries, starting from the first zero-crossing point, the waveform is truncated to the next zero-crossing point in the same direction to form a complete waveform segment. Subsequent waveforms are truncated in the same way to ensure that each segment contains a complete waveform fluctuation process. All the truncated segments together constitute the current signal periodic segment of the standardized time-domain current waveform.

[0148] The DC component of each current sampling point in the current signal period is extracted one by one. These DC components are constant parts of the sampling points that do not change with time. The DC components of all sampling points are summarized, and the average value of these DC components is calculated to obtain the DC bias of the current signal period that can represent the DC offset level of the entire current signal period.

[0149] The obtained DC bias is used as the correction reference. Each current sampling point in the current signal periodic segment is processed one by one. The DC bias is subtracted from the original amplitude of each sampling point to eliminate the influence of DC bias on the waveform. This allows the corrected waveform to be symmetrically distributed around the zero axis, and finally, the corrected periodic current signal of the standardized time-domain current waveform is obtained.

[0150] A comprehensive analysis of the amplitude distribution of the current signal in the correction period is conducted. The peak amplitude, amplitude distribution range, and number of sampling points in different amplitude intervals above and below the zero axis are compared. The degree of difference in amplitude distribution on both sides is calculated. This degree of difference is quantified and integrated to obtain the waveform distortion characteristic quantity that reflects the amplitude distribution of the waveform deviating from the symmetrical shape. This waveform distortion characteristic quantity is the higher-order statistical moment of the normalized signal matrix.

[0151] The complete law of amplitude change of the correction period current signal over time is continuously tracked. By decomposing the signal in the time domain into a series of sinusoidal components of different frequencies, the amplitude and phase information of each frequency component are captured one by one. These frequency-related information are arranged in an orderly manner from low to high frequency to form the period spectrum of the correction period current signal that can intuitively reflect the frequency composition of the signal.

[0152] The periodic spectrum is continuously divided according to the frequency range to form multiple continuous and non-overlapping frequency bands. All spectral components contained in each frequency band are extracted one by one, and the square value of the amplitude of each spectral component is calculated. Then, the square values ​​of the amplitudes of all spectral components in the same frequency band are accumulated. The accumulated result is the basic energy of the frequency band.

[0153] The basic energy values ​​corresponding to all the divided frequency bands are summarized, and the basic energy of each frequency band is accumulated in turn to ensure that the energy contribution of any frequency band is not missed. The total value obtained by the complete accumulation calculation is the total energy of the periodic spectrum.

[0154] The ratio of the base energy to the total energy of each frequency band is calculated. All ratios are arranged in order of frequency of the corresponding frequency band to form the distribution ratio vector of the total energy. By assigning different weight coefficients to different ratio values, the contribution of the ratio values ​​of key frequency bands is highlighted, while the interference of the ratio values ​​of secondary frequency bands is suppressed. The weighted ratio vector is integrated and processed to finally obtain the energy distribution characteristics of the standardized signal matrix.

[0155] In the calculation of energy distribution characteristics, The value representing the total number of frequency bands comes from the periodic spectrum obtained after performing a spectral transformation on the current signal of the correction periodic segment. This value is obtained by dividing the periodic spectrum into continuous and non-overlapping frequency ranges.

[0156] The value representing the basic energy of a single frequency band comes from the spectral components within the corresponding frequency band. For all spectral components contained in each frequency band, the square of their amplitude is calculated and accumulated. The accumulated result is the basic energy of that frequency band.

[0157] The value representing the stability coefficient is a fixed value set in advance. Its purpose is to prevent abnormal calculation results due to certain extreme situations in subsequent calculations, and to ensure the stability of the entire calculation process.

[0158] The significance of this calculation lies in the quantitative integration of the energy distribution of the periodic spectrum. By processing the ratio of the basic energy of each frequency band to the total energy, and fusing the energy distribution information of all frequency bands, the energy distribution characteristics that can accurately reflect the spectral changes in the standardized signal matrix are finally obtained. These characteristics can clearly show the contribution ratio of different frequency bands to the total energy, providing a reliable basis for the subsequent construction of holographic feature vectors.

[0159] By tracking the instantaneous amplitude changes and frequency fluctuations of the current signal during the correction period, the time-domain information and frequency-domain information of the signal are dynamically correlated. By decomposing the instantaneous frequency components and corresponding amplitudes of the signal at different times, while retaining the continuous characteristics of the frequency components at each time, these correlated information are arranged in an orderly manner according to time sequence and frequency level, and a two-dimensional data array is constructed with rows and columns corresponding to the time dimension and frequency dimension, respectively, to obtain the time-frequency distribution matrix of the current signal during the correction period.

[0160] Based on the time-frequency plane corresponding to the time-frequency distribution matrix, continuous time intervals are divided along the time axis with a fixed time span, and continuous frequency intervals are divided along the frequency axis with a fixed frequency span. The time intervals and frequency intervals intersect to form multiple rectangular regions with clear boundaries and no overlap. Each rectangular region corresponds to a specific time range and frequency range. These rectangular regions are the rectangular analysis areas of the mining variable frequency speed control system.

[0161] Each rectangular analysis region is selected one by one, and all time-frequency components contained in the region are extracted. The energy value of each time-frequency component is determined by the magnitude of its corresponding time. The energy values ​​of all time-frequency components in the same rectangular analysis region are continuously accumulated to ensure that the energy value of each component is included in the calculation range. The total value after accumulation is the total energy of the rectangular analysis region.

[0162] The total energy of all rectangular analysis regions is summarized, and the ratio of the total energy of each rectangular analysis region to the sum of the total energy of all regions is calculated to obtain the energy proportion of each region. By performing a comprehensive quantitative analysis of all energy proportions, the uniformity of the distribution of time-frequency energy in different rectangular analysis regions is measured, and finally the energy entropy of the standardized signal matrix, which can reflect the time-frequency energy distribution characteristics of the standardized signal matrix, is obtained.

[0163] The value representing the total number of rectangular analysis regions comes from the time-frequency distribution matrix obtained after joint time-frequency reconstruction of the current signal in the correction period. On the time-frequency plane, rectangular analysis regions with clear boundaries and no overlap are divided according to fixed time span and frequency span. The total number of all the rectangular analysis regions formed by the division is this value.

[0164] The value representing the energy percentage of a single rectangular analysis region is derived from the calculation of the total energy of the corresponding rectangular analysis region and the sum of the total energy of all rectangular analysis regions. First, the energy values ​​of all time-frequency components in each rectangular analysis region are accumulated to obtain the total energy of each region. Then, the total energy of each region is divided by the sum of the total energy of all regions. The result is the proportion of the energy of the rectangular analysis region to the total energy of the region.

[0165] The significance of this calculation lies in quantifying the energy proportion of each rectangular analysis area, comprehensively processing all proportions, accurately measuring the uniformity of time-frequency energy distribution in different time and frequency ranges, and finally obtaining the energy entropy that can comprehensively reflect the energy distribution characteristics of the time-frequency components of the standardized signal matrix. This provides a reliable basis for the construction of subsequent holographic feature vectors and helps improve the accuracy of fault diagnosis.

[0166] The beneficial effects are as follows: by identifying the periodic zero-crossing points of the standardized time-domain current waveform, segmenting the periodic segments of the current signal, aggregating the DC component to obtain the DC bias and performing bias correction, and then analyzing the amplitude distribution asymmetry to extract higher-order statistical moments, the current signal of the correction period segment is subjected to spectral transformation, the basic energy of the frequency band and the overall energy are determined and nonlinearly fused to obtain the energy distribution characteristics. Furthermore, by jointly reconstructing the time and frequency, dividing the rectangular analysis area, and accumulating the overall energy of the area to analyze the energy entropy, the core features of the standardized signal matrix in the time domain, frequency domain and time-frequency domain are comprehensively captured. This ensures that the extracted features can accurately reflect the waveform distortion, energy distribution and time-frequency distribution characteristics of the signal, providing comprehensive and reliable multi-dimensional feature support for the subsequent construction of holographic feature vectors, and effectively improving the accuracy and effectiveness of fault diagnosis of mining variable frequency speed control systems.

[0167] The low-dimensional feature module 104 is used to map the holographic feature vector to a preset feature importance evaluation network, and rearrange the features in the holographic feature vector whose feature dimension weight is lower than the standard threshold in the preset feature importance evaluation network to obtain a low-dimensional feature subset of the mining variable frequency speed control system.

[0168] In this embodiment of the invention, when the low-dimensional feature module performs the operation of mapping the holographic feature vector to a preset feature importance evaluation network and rearranging the features whose feature dimension weights in the holographic feature vector are lower than the standard threshold in the preset feature importance evaluation network to obtain the low-dimensional feature subset of the mining variable frequency speed control system, it is specifically used for:

[0169] Based on a preset feature importance evaluation network, the importance of the holographic feature vector is evaluated to obtain the initial weight values ​​of the feature dimensions in the holographic feature vector;

[0170] The initial weight values ​​of the feature dimensions are dynamically adjusted to obtain the adjusted weight values ​​of the feature dimensions.

[0171] Based on the overall distribution characteristics of the corrected weight values, the weight screening threshold of the preset feature importance evaluation network is determined;

[0172] The corrected weight value is mapped to the weight screening threshold to filter out low-dimensional features whose feature dimension is lower than the weight screening threshold, thereby obtaining the feature dimension to be reconstructed in the holographic feature vector;

[0173] Analyze the intrinsic correlation between the feature dimension to be reconstructed and the high-weight feature dimension whose corrected weight value is not lower than the weight screening threshold;

[0174] Based on the inherent correlation, the feature dimension to be reconstructed is directionally fused to the high-weight feature dimension to obtain a low-dimensional feature subset of the mining variable frequency speed control system.

[0175] The holographic feature vector is input into a preset feature importance evaluation network. This network is constructed in advance based on the fault diagnosis requirements and feature correlation rules of the mining variable frequency speed control system. It includes the corresponding correlation rules between feature dimensions and fault types and fault degrees. By comparing the attributes and performance of each feature dimension in the holographic feature vector with the correlation rules in the network one by one, and combining the influence depth and correlation of the feature dimension on fault judgment, a corresponding fixed value is assigned to each feature dimension. This value is the initial weight value of the feature dimension in the holographic feature vector.

[0176] Collect the initial weight values ​​and corresponding diagnostic effects of the same feature dimension in the historical fault diagnosis data of the mining variable frequency speed control system. Compare the difference between the initial weight value of the current feature dimension and the feature weight distribution of effective diagnostic cases in the historical data. Analyze whether the performance of the current feature dimension in the actual signal is consistent with the basis for assigning the initial weight value. Make targeted adjustments to the initial weight values ​​with deviations to eliminate weight distortion caused by accidental factors and obtain the corrected weight value of the feature dimension.

[0177] A comprehensive statistical analysis of the corrected weight values ​​for all feature dimensions is conducted to determine the concentration range, distribution density, and peak position of these values. This clarifies the overall distribution pattern of the corrected weight values ​​and identifies the numerical boundary points that clearly distinguish between core and secondary features. These boundary points ensure that feature dimensions with values ​​above this threshold play a crucial supporting role in fault diagnosis, while feature dimensions with values ​​below this threshold have limited impact on the diagnostic results. This boundary point is the weight screening threshold for the preset feature importance assessment network.

[0178] The corrected weight value of each feature dimension is compared with the determined weight screening threshold one by one to determine whether each corrected weight value is lower than the threshold. All feature dimensions with corrected weight values ​​lower than the weight screening threshold are classified as low-dimensional features. The set of these low-dimensional features is integrated to obtain the feature dimensions to be reconstructed in the holographic feature vector.

[0179] Starting from the signal characteristics of the mine variable frequency speed control system reflected by the feature dimensions, the correlation path with the fault type, and the indication role of the system operating status, we analyze the relationship between the feature dimensions to be reconstructed and the high-weight feature dimensions whose corrected weight values ​​are not lower than the weight screening threshold. We find the common direction and complementary relationship between the two in describing the system operating status, locating the fault location, and judging the severity of the fault, and clarify the inherent correlation logic between the two.

[0180] Based on the inherent correlation logic derived from the analysis, the effective information contained in the feature dimension to be reconstructed is directionally integrated into the corresponding high-weight feature dimension to supplement the deficiencies of the high-weight feature dimension in terms of detailed description and scope coverage. At the same time, duplicate and redundant information that occurs during the fusion process is eliminated to ensure that the integrated features retain the core key information while achieving dimensional simplification, and finally obtain the low-dimensional feature subset of the mining variable frequency speed control system.

[0181] The beneficial effects are as follows: by using a preset feature importance evaluation network to evaluate the importance of holographic feature vectors and dynamically adjust the feature dimension weights, the accuracy and adaptability of weight assignment are ensured. Based on the distribution characteristics of the adjusted weights, a reasonable weight screening threshold is determined, and low-dimensional features are accurately screened to form the feature dimensions to be reconstructed. By analyzing the intrinsic correlation between these features and high-weight feature dimensions, targeted fusion is performed. While simplifying feature dimensions and eliminating redundant information, the core features that are key to fault diagnosis are fully retained. The resulting low-dimensional feature subset reduces the processing load of subsequent fault diagnosis and ensures the effectiveness and relevance of the features, providing solid support for improving the efficiency and accuracy of fault diagnosis in mining variable frequency speed control systems.

[0182] The fault diagnosis module 105 is used to evaluate the confidence level of the low-dimensional feature subset and perform decision fusion on the evaluation results to obtain the preliminary fault diagnosis of the mining variable frequency speed control system.

[0183] In this embodiment of the invention, when the fault diagnosis module performs confidence assessment on the low-dimensional feature subset and performs decision fusion on the assessment results to obtain a preliminary fault diagnosis of the mining variable frequency speed control system, it is specifically used for:

[0184] The matching degree analysis is performed between the simplified features in the low-dimensional feature subset and the indication strength of the preset fault type to obtain the first layer confidence of the low-dimensional feature subset;

[0185] The first layer confidence score is subjected to time smoothing correction to obtain the second layer confidence score of the low-dimensional feature subset;

[0186] The first layer confidence and the second layer confidence are jointly corrected to obtain the target feature confidence of the low-dimensional feature subset;

[0187] Based on the confidence level of the target features, determine the comprehensive probability of state-related faults of the mining variable frequency speed control system;

[0188] The overall probability of the state fault is mapped to the preset fault type to obtain the preliminary fault diagnosis of the mining variable frequency speed control system.

[0189] Extract all simplified features from the low-dimensional feature subset. These features are the core fault-related features retained after dimensionality reconstruction and redundancy removal. At the same time, retrieve the indication intensity corresponding to the preset fault type. This indication intensity is a standard reference set in advance based on the typical characteristics of various faults. Compare the attributes, performance and association logic of each simplified feature with the indication intensity of each preset fault type one by one. Quantify the degree of consistency between the two. The higher the degree of consistency, the higher the fixed value is assigned. The integration of the quantification results corresponding to all simplified features is the first layer confidence of the low-dimensional feature subset.

[0190] Collect all values ​​of the first-level confidence score within a continuous time range, track their changing trends over time, identify abrupt changes in values ​​caused by instantaneous signal fluctuations, use the confidence score values ​​at adjacent time points as a reference benchmark, and make appropriate adjustments to the abrupt changes in values ​​according to the principle of the stability of numerical changes to eliminate the bias caused by instantaneous interference, so that the adjusted confidence score values ​​can reflect the stable state over a continuous time period. The adjusted result is the second-level confidence score of the low-dimensional feature subset.

[0191] This study analyzes the immediate matching accuracy of the first-level confidence level and the time stability advantage of the second-level confidence level, clarifies their complementary roles in fault diagnosis, and integrates the values ​​of the two levels of confidence. Based on the principle of reasonable compromise between the two levels of values, it retains the accurate reflection of the current feature matching by the first-level confidence level while incorporating the time-stationary characteristics of the second-level confidence level. Through the organic fusion of values, the limitations of a single confidence level are eliminated, and the target feature confidence of the low-dimensional feature subset is finally obtained.

[0192] Summarize the confidence scores of the target features corresponding to all simplified features, comprehensively analyze the overall distribution and central tendency of these confidence scores, determine the degree to which they point to the fault state, combine the typical distribution patterns of feature confidence scores when various faults occur, quantify the overall degree of pointing to the fault state, and form a clear result that can comprehensively reflect whether the system has a fault and the probability of the fault. This result is the comprehensive probability of the state fault of the mining variable frequency speed control system.

[0193] The fault probability range standard corresponding to all preset fault types is retrieved, and the comprehensive probability of the obtained state fault is compared with the probability range of each preset fault type one by one. The preset fault type that perfectly matches the comprehensive probability is found, and the preset fault type is determined as the most likely fault in the system. This determined fault type result is the preliminary fault diagnosis of the mining variable frequency speed control system.

[0194] The beneficial effects are as follows: by performing a matching degree analysis between the simplified features in the low-dimensional feature subset and the indication intensity of the preset fault type to obtain the first layer of confidence, and combining it with time smoothing correction to eliminate the interference caused by instantaneous signal fluctuations to obtain a more stable second layer of confidence, and then integrating the advantages of the two layers of confidence through collaborative correction to obtain the target feature confidence, the comprehensive probability of the state fault of the mining variable frequency speed control system can be determined in a comprehensive and accurate manner based on the target feature confidence. This is then mapped to the preset fault type, which not only ensures the accuracy and pertinence of fault judgment, but also improves the stability of diagnostic results. It provides reliable intermediate support for the subsequent determination of the final fault type, severity and location, and effectively helps to promote the efficient advancement of the overall fault diagnosis process.

[0195] In this embodiment of the invention, the diagnostic report module 106 is used to perform a time-series comparison between the preliminary fault diagnosis and the historical health status baseline of the mining variable frequency speed control system to obtain the final fault type, severity and possible location information of the mining variable frequency speed control system, so as to generate a structured diagnostic report of the mining variable frequency speed control system.

[0196] The historical health status baseline of the mining variable frequency speed control system is retrieved. This baseline is a collection of standard ranges and variation patterns of various operating parameters, signal characteristics, and status performance accumulated during long-term healthy operation of the system. It includes core reference indicators related to fault diagnosis. All relevant data of this baseline are completely extracted from the system's storage unit to ensure the integrity and accuracy of the data.

[0197] The fault-related features included in the preliminary fault diagnosis are arranged in chronological order, and the data of the historical health status baseline are also sorted out according to the corresponding time dimension, so that the two maintain a consistent comparison rhythm on the time axis. The features of the preliminary fault diagnosis are compared with the standard features of the corresponding time period of the baseline in each time period. The deviations between the two in terms of parameter values, changing trends, and feature performance are analyzed to clarify the specific content and manifestation of the deviations.

[0198] Based on the deviation characteristics obtained from the time-series comparison, the deviation characteristics are compared with the typical characteristics library of various known fault types of mining variable frequency speed control systems. The degree of match between the deviation characteristics and the typical fault characteristics is matched one by one. The fault types that do not match are eliminated, and the fault types that completely match the deviation characteristics are determined. This fault type is the final fault type of the mining variable frequency speed control system.

[0199] By analyzing the magnitude, duration, and impact of deviation characteristics on key functions of normal system operation in the timing comparison, and combining the normal fluctuation range of such parameters in the historical health status baseline, the severity of the impact of the fault on system operation is determined, a clear severity level is defined, and information on the severity of faults in the mining variable frequency speed control system is generated.

[0200] Based on the typical manifestations and impact range of the final fault type, combined with the structural composition, component functions and signal transmission paths of the mining variable frequency speed control system, the system components or areas corresponding to the fault characteristics are analyzed to determine the most likely specific location of the fault. At the same time, referring to the location experience of similar faults in historical fault cases, the rationality of the location results is further verified, and possible location information of the fault in the mining variable frequency speed control system is obtained.

[0201] The final fault type, severity, and possible location information are systematically organized and arranged in a logical order of fault overview, feature analysis, diagnostic conclusion, and location description. The fault-related time sequence comparison basis and reference standards are supplemented to form a document with a clear structure, complete content, and standardized expression. This document is the structured diagnostic report of the mining variable frequency speed control system.

[0202] The beneficial effects are that by comparing the preliminary fault diagnosis with the historical health status baseline of the mining variable frequency speed control system in a time series, the reference value of historical health operation data is fully utilized to accurately verify the rationality and accuracy of the preliminary diagnosis results, effectively eliminate misjudgments caused by transient interference, and clearly identify the final fault type, severity and possible location information of the system. The structured diagnostic report generated based on this core information is well-structured and detailed, comprehensively presenting key details related to the fault. This provides an intuitive and reliable basis for staff to quickly grasp the fault situation and formulate targeted handling plans, further ensuring the timeliness and effectiveness of fault handling in the mining variable frequency speed control system.

[0203] Reference Figure 2 The diagram shown is a flowchart illustrating a fault diagnosis method for a mine-use variable frequency speed control system according to an embodiment of the present invention. In this embodiment, the fault diagnosis method for the mine-use variable frequency speed control system includes:

[0204] S1. Synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system, and integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system;

[0205] S2. Filter the noise in different dimensions of the original multi-source heterogeneous signal set, and reconstruct the dimensions of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system;

[0206] S3. Extract the higher-order statistical moments of the time-domain waveform in the standardized signal matrix, evaluate the energy distribution characteristics of the spectral changes in the standardized signal matrix, and analyze the energy entropy of the time-frequency components in the standardized signal matrix to construct the holographic feature vector of the mining variable frequency speed control system;

[0207] S4. Map the holographic feature vector to a preset feature importance evaluation network, and rearrange the features in the holographic feature vector whose feature dimension weight is lower than the standard threshold in the preset feature importance evaluation network to obtain a low-dimensional feature subset of the mining variable frequency speed control system;

[0208] S5. Confidence assessment is performed on the low-dimensional feature subset, and decision fusion is performed on the assessment results to obtain the preliminary fault diagnosis of the mining variable frequency speed control system.

[0209] S6. The preliminary fault diagnosis is compared with the historical health status baseline of the mining variable frequency speed control system in a time series to obtain the final fault type, severity and possible location information of the mining variable frequency speed control system, so as to generate a structured diagnostic report of the mining variable frequency speed control system.

[0210] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0211] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0212] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A fault diagnosis system for a mine variable frequency speed control system, characterized in that, The system includes an information acquisition module, a matrix construction module, a high-dimensional vector module, a low-dimensional feature module, a fault diagnosis module, and a diagnostic report module, wherein: The information acquisition module is used to synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system, so as to integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system. The matrix construction module is used to filter noise of different dimensions in the original multi-source heterogeneous signal set, and to reconstruct the dimensions of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system. The high-dimensional vector module is used to extract the higher-order statistical moments of the time-domain waveform in the standardized signal matrix, evaluate the energy distribution characteristics of the spectral changes in the standardized signal matrix, and analyze the energy entropy of the time-frequency components in the standardized signal matrix to construct the holographic feature vector of the mining variable frequency speed control system. The low-dimensional feature module is used to map the holographic feature vector to a preset feature importance evaluation network, and to rearrange the features in the holographic feature vector whose feature dimension weight is lower than the standard threshold in the preset feature importance evaluation network to obtain a low-dimensional feature subset of the mining variable frequency speed control system. The fault diagnosis module is used to evaluate the confidence level of the low-dimensional feature subset and perform decision fusion on the evaluation results to obtain the preliminary fault diagnosis of the mining variable frequency speed control system. The diagnostic report module is used to compare the preliminary fault diagnosis with the historical health status baseline of the mining variable frequency speed control system in a time series to obtain the final fault type, severity and possible location information of the mining variable frequency speed control system, so as to generate a structured diagnostic report of the mining variable frequency speed control system.

2. The fault diagnosis system for a mining variable frequency speed control system as described in claim 1, characterized in that, When the information acquisition module synchronously acquires the three-phase output current, DC bus voltage, and power device temperature signals of the mining variable frequency speed control system to integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system, it is specifically used for: Synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system to obtain the multi-source signals of the mining variable frequency speed control system; The DC bus voltage is used as the time base synchronization reference signal for the multi-source signal, and the target sampling interval of the mining variable frequency speed control system is determined based on the period of the time base synchronization reference signal. Based on the target sampling interval, the original discrete sequence of the three-phase output current is resampled using time-base synchronization to obtain the synchronized three-phase output current of the mining variable frequency speed control system. The synchronized three-phase output current, the time base synchronization reference signal, and the power device temperature signal are time-aligned to obtain the original multi-source heterogeneous signal set of the mining variable frequency speed control system.

3. The fault diagnosis system for a mining variable frequency speed control system as described in claim 1, characterized in that, When the matrix construction module performs filtering of noise in different dimensions from the original multi-source heterogeneous signal set and reconstructs the dimensions of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system, it is specifically used for: Based on the signal change rate of the mining variable frequency speed control system, the signals in the original multi-source heterogeneous signal set are divided into fast-changing dimension signals and slow-changing dimension signals. Based on the amplitude change between adjacent sampling points in the rapidly changing dimension signal, a first dynamic filtering threshold for the rapidly changing dimension signal is set. Signal points in the rapidly changing dimensional signal whose amplitude changes are lower than the first dynamic filtering threshold are treated as high-frequency background noise and filtered out to obtain the first filtered signal of the original multi-source heterogeneous signal set. Within a preset time window, a second dynamic filtering threshold for the slowly varying dimension signal is set based on the fluctuation range of the slowly varying dimension signal. Signal segments whose fluctuation range is consistently below the second dynamic filtering threshold are treated as low-frequency drift noise and filtered out to obtain the second filtered signal of the slowly varying dimension signal.

4. The fault diagnosis system for a mining variable frequency speed control system as described in claim 3, characterized in that, When the matrix construction module performs dimensional reconstruction on the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system, it is specifically used for: The first filtered signal and the second filtered signal are bidirectionally coupled to obtain the standard filtered signal of the mining variable frequency speed control system. The standard filtered signal is mapped to the multi-dimensional data space of the mining variable frequency speed control system; Within the multidimensional data space, the aligned signal segment sequence of the standard filtered signal is obtained based on the relative proportions of the alignment and amplitude range of the signal segments on the time axis in the standard filtered signal. The amplitude normalization process is performed on the aligned signal segment sequence to obtain the normalized signal sequence of the mining variable frequency speed control system; The normalized signal sequence is reconstructed to obtain the standardized signal matrix of the mining variable frequency speed control system.

5. The fault diagnosis system for a mining variable frequency speed control system as described in claim 1, characterized in that, When the high-dimensional vector module extracts the higher-order statistical moments of the time-domain waveform from the standardized signal matrix, it is specifically used for: Identify the periodic zero-crossing points of the standardized time-domain current waveform in the standardized signal matrix; Using the periodic zero-crossing point as the dividing criterion, the standardized time-domain current waveform is divided into current signal period segments of the standardized time-domain current waveform; By aggregating the DC components of the current sampling points in the current signal periodic segment, the DC bias of the current signal periodic segment is obtained; Based on the DC bias, the current sampling point is biased and corrected to obtain the corrected period current signal of the standardized time-domain current waveform. The asymmetry of amplitude distribution in the current signal of the correction period is analyzed to obtain the waveform distortion characteristics of the amplitude distribution, which are used as the higher-order statistical moments of the standardized signal matrix.

6. The fault diagnosis system for a mining variable frequency speed control system as described in claim 5, characterized in that, The high-dimensional vector module, when evaluating the energy distribution characteristics of spectral variations in the normalized signal matrix, is specifically used for: The current signal of the correction period segment is subjected to spectral transformation to obtain the period segment spectrum of the current signal of the correction period segment; The fundamental energy of the frequency band is determined based on the sum of the squares of the amplitudes of the spectral components in the frequency band of the periodic segment spectrum. The basic energy is integrated into the total energy of the periodic spectrum; The distribution ratio vector of the overall energy is nonlinearly fused to obtain the energy distribution characteristics of the standardized signal matrix, wherein the calculation formula of the energy distribution characteristics is as follows: ； In the formula, The energy distribution characteristics are as described above. It is an exponential function. The total number of the frequency bands. For the first The fundamental energy of each frequency band, The preset stability coefficient, It is a logarithmic function.

7. The fault diagnosis system for a mining variable frequency speed control system as described in claim 6, characterized in that, When performing the analysis of the energy entropy of the time-frequency components in the normalized signal matrix, the high-dimensional vector module is specifically used for: The current signal of the correction period segment is subjected to joint time-frequency reconstruction to obtain the time-frequency distribution matrix of the current signal of the correction period segment; On the time-frequency plane, the time-frequency distribution matrix is ​​divided into a rectangular analysis region for the mining variable frequency speed control system; The total energy of the rectangular analysis area is obtained by summing the energy values ​​of all time-frequency components within the rectangular analysis area. Based on the total energy of the region, the energy entropy of the standardized signal matrix is ​​analyzed, wherein the formula for calculating the energy entropy is as follows: ； In the formula, The energy entropy, This represents the total number of rectangular analysis regions. For the first The proportion of energy in each rectangular analysis region to the total energy of the region. It is a logarithmic function.

8. The fault diagnosis system for a mining variable frequency speed control system as described in claim 1, characterized in that, When the low-dimensional feature module performs the operation of mapping the holographic feature vector to a preset feature importance evaluation network and rearranging the features whose feature dimension weights in the holographic feature vector are lower than the standard threshold in the preset feature importance evaluation network to obtain the low-dimensional feature subset of the mining variable frequency speed control system, it is specifically used for: Based on a preset feature importance evaluation network, the importance of the holographic feature vector is evaluated to obtain the initial weight values ​​of the feature dimensions in the holographic feature vector; The initial weight values ​​of the feature dimensions are dynamically adjusted to obtain the adjusted weight values ​​of the feature dimensions. Based on the overall distribution characteristics of the corrected weight values, the weight screening threshold of the preset feature importance evaluation network is determined; The corrected weight value is mapped to the weight screening threshold to filter out low-dimensional features whose feature dimension is lower than the weight screening threshold, thereby obtaining the feature dimension to be reconstructed in the holographic feature vector; Analyze the intrinsic correlation between the feature dimension to be reconstructed and the high-weight feature dimension whose corrected weight value is not lower than the weight screening threshold; Based on the inherent correlation, the feature dimension to be reconstructed is directionally fused to the high-weight feature dimension to obtain a low-dimensional feature subset of the mining variable frequency speed control system.

9. The fault diagnosis system for a mining variable frequency speed control system as described in claim 1, characterized in that, When the fault diagnosis module performs confidence assessment on the low-dimensional feature subset and performs decision fusion on the assessment results to obtain a preliminary fault diagnosis of the mining variable frequency speed control system, it is specifically used for: The matching degree analysis is performed between the simplified features in the low-dimensional feature subset and the indication strength of the preset fault type to obtain the first layer confidence of the low-dimensional feature subset; The first layer confidence score is subjected to time smoothing correction to obtain the second layer confidence score of the low-dimensional feature subset; The first layer confidence and the second layer confidence are jointly corrected to obtain the target feature confidence of the low-dimensional feature subset; Based on the confidence level of the target features, determine the comprehensive probability of state-related faults of the mining variable frequency speed control system; The overall probability of the state fault is mapped to the preset fault type to obtain the preliminary fault diagnosis of the mining variable frequency speed control system.

10. A fault diagnosis method for a mine variable frequency speed control system, characterized in that, The method is used in the fault diagnosis system for the mining variable frequency speed control system according to claim 1. S1. Synchronously acquire the three-phase output current, DC bus voltage and power device temperature signals of the mining variable frequency speed control system, and integrate them into the original multi-source heterogeneous signal set of the mining variable frequency speed control system; S2. Filter the noise in different dimensions of the original multi-source heterogeneous signal set, and reconstruct the dimensions of the filtered result to obtain the standardized signal matrix of the mining variable frequency speed control system; S3. Extract the higher-order statistical moments of the time-domain waveform in the standardized signal matrix, evaluate the energy distribution characteristics of the spectral changes in the standardized signal matrix, and analyze the energy entropy of the time-frequency components in the standardized signal matrix to construct the holographic feature vector of the mining variable frequency speed control system; S4. Map the holographic feature vector to a preset feature importance evaluation network, and rearrange the features in the holographic feature vector whose feature dimension weight is lower than the standard threshold in the preset feature importance evaluation network to obtain a low-dimensional feature subset of the mining variable frequency speed control system; S5. Confidence assessment is performed on the low-dimensional feature subset, and decision fusion is performed on the assessment results to obtain the preliminary fault diagnosis of the mining variable frequency speed control system. S6. The preliminary fault diagnosis is compared with the historical health status baseline of the mining variable frequency speed control system in a time series to obtain the final fault type, severity and possible location information of the mining variable frequency speed control system, so as to generate a structured diagnostic report of the mining variable frequency speed control system.

Images