A fault diagnosis method and system of an automatic tension control valve of a flight conveyor

Abstract

The application relates to the technical field of mechanical equipment fault diagnosis, and discloses a fault diagnosis method and system of an automatic tension control valve of a scraper conveyor; the method comprises the following steps: collecting pressure and flow data of a main flow channel and an auxiliary flow channel through a sensor array, extracting frequency domain features to obtain a standardized frequency spectrum; detecting parameter coupling abnormalities and generating abnormal indexes based on a frequency spectrum cross-correlation coefficient; fusing environmental interference and medium viscosity data, obtaining nonlinear feature distribution and correction deviation representing mechanical states through classification and filtering analysis; matching the correction deviation with a preset fault mode, determining a fault type, and generating fault positioning coordinates in a three-dimensional space by using an optimization algorithm; finally, verifying the stability of the coordinates through fluid dynamics simulation of a high-fidelity virtual model, and outputting a precise early warning when the conditions are met; the application improves the accuracy of early internal fault diagnosis of a complex double-flow valve through multi-source information fusion and intelligent algorithm cooperation.

Description

Technical Field

[0001] This application relates to the field of mechanical equipment fault diagnosis technology, and in particular to a fault diagnosis method and system for an automatic tension control valve of a scraper conveyor. Background Technology

[0002] Scraper conveyors are widely used in material transportation in heavy industries such as coal and ore. Their automatic tension control valves play a crucial role in ensuring stable conveyor operation. The automatic tension control valves regulate the hydraulic systems of the main flow channel and auxiliary flow channels to ensure the equipment maintains tension under high load conditions. However, during long-term operation, problems such as valve core jamming and seal wear may occur. These faults not only affect the equipment's working efficiency but may also lead to serious damage, thus impacting production safety and economic benefits. Currently, fault diagnosis for automatic tension control valves mainly relies on traditional valve pressure detection or flow monitoring methods. These methods typically only monitor parameters of the main flow channel or auxiliary flow channel individually and cannot effectively reflect the coupling effect between the two channels. Due to the complex nonlinear characteristics of hydraulic flow within the flow channel system and its influence from external factors such as temperature and medium viscosity, traditional fault diagnosis methods struggle to accurately identify and locate faults in the system, especially under complex operating conditions, often resulting in misdiagnosis or missed diagnosis.

[0003] To address the aforementioned issues, this application provides a fault diagnosis method for automatic tension control valves based on a combination of signal processing, classification algorithms, and dynamic simulation. By acquiring pressure and flow data in real time and combining advanced spectrum analysis and nonlinear feature analysis techniques, this method can more accurately identify and locate faults in the automatic tension control valve system, thereby improving the accuracy and real-time performance of fault diagnosis and ensuring the safe and stable operation of the equipment. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a fault diagnosis method and system for the automatic tension control valve of a scraper conveyor, enabling accurate, early, and reliable diagnosis of internal faults in the automatic tension control valve of the scraper conveyor.

[0005] In a first aspect, this application provides a fault diagnosis method for an automatic tension control valve of a scraper conveyor, the method comprising:

[0006] Step S1: Collect pressure and flow data of the main channel and auxiliary channel of the automatic tensioning control valve through a sensor array, and extract the frequency domain features of the pressure and flow data using signal processing methods to obtain the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel.

[0007] Step S2: Calculate the cross-correlation coefficient based on the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel, and determine whether the cross-correlation coefficient is lower than a preset threshold. If so, determine that there is a parameter coupling anomaly between the main channel and the auxiliary channel, and generate an abnormal coupling index.

[0008] Step S3: Analyze the abnormal coupling index by combining the environmental interference data of the scraper conveyor to obtain the initial distribution range of the nonlinear flow characteristics. Correct the initial distribution range by integrating the medium characteristic data and determine whether the corrected distribution range deviates from the normal reference. If so, output the correction deviation value.

[0009] Step S4: Calculate the similarity score between the correction deviation value and the preset fault mode. If the similarity score is higher than the preset threshold, determine the corresponding fault type. Based on the similarity score, generate fault location coordinates through optimization and mapping processing.

[0010] Step S5: Map the fault location coordinates to the virtual model of the automatic tension control valve, and verify the stability of the fault location coordinates through dynamic simulation. If the stability index meets the preset conditions, generate a real-time warning signal containing the fault type and the fault location coordinates.

[0011] Secondly, this application provides a fault diagnosis system for an automatic tension control valve of a scraper conveyor, the system comprising:

[0012] The data acquisition module is used to acquire pressure and flow data of the main channel and auxiliary channel of the automatic tension control valve through a sensor array, and to extract the frequency domain features of the pressure and flow data using signal processing methods to obtain the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel.

[0013] The anomaly analysis module is used to calculate the cross-correlation coefficient based on the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel, determine whether the cross-correlation coefficient is lower than a preset threshold, and if so, determine that there is a parameter coupling anomaly between the main channel and the auxiliary channel, and generate an anomaly coupling index.

[0014] The correction processing module is used to analyze the abnormal coupling index by combining the environmental interference data of the scraper conveyor, obtain the initial distribution range of the nonlinear flow characteristics, correct the initial distribution range by integrating the medium characteristic data, and determine whether the corrected distribution range deviates from the normal reference. If so, the correction deviation value is output.

[0015] The preliminary positioning module is used to calculate the similarity score between the correction deviation value and the preset fault mode. If the similarity score is higher than the preset threshold, the corresponding fault type is determined. Based on the similarity score, fault positioning coordinates are generated through optimization and mapping processing.

[0016] The stability verification module is used to map the fault location coordinates to the virtual model of the automatic tension control valve, and verify the stability of the fault location coordinates through dynamic simulation. If the stability index meets the preset conditions, a real-time warning signal containing the fault type and the fault location coordinates is generated.

[0017] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0018] This application provides a fault diagnosis method and system for an automatic tension control valve of a scraper conveyor. First, a high-frequency synchronous sensor array is used to collect pressure and flow data of the dual-channel system, and a standardized frequency domain spectrum is extracted by Fourier transform, providing a high signal-to-noise ratio characteristic basis for subsequent analysis. By calculating the cross-correlation coefficient between the spectra and judging coupling anomalies based on statistical thresholds, the system can keenly capture the destruction of the dynamic coordination of the dual-channel system caused by early faults such as valve core jamming and seal wear. Furthermore, the abnormal indicators reflecting the internal coupling state are fused and analyzed with environmental temperature interference, and a support vector machine is used to identify nonlinear flow characteristics and their statistical distribution laws that characterize different fault precursors from complex data.

[0019] This application also corrects the feature distribution by dynamically fusing medium viscosity data using Kalman filtering, effectively eliminating the interference of operating condition changes on diagnostic features and focusing the diagnostic attention on the health status of the mechanical components themselves. Subsequently, the corrected features are matched with preset fault modes for similarity, and combined with genetic algorithm optimization and principal component analysis for dimensionality reduction mapping, realizing the transformation from abstract similarity to precise fault location coordinates in the three-dimensional space of the valve body. Finally, by performing dynamic simulation based on computational fluid dynamics in a high-fidelity virtual model, the physical rationality and stability of the location coordinates are verified, ensuring the reliability of the diagnostic conclusions. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a fault diagnosis method for an automatic tension control valve of a scraper conveyor according to an embodiment of this application.

[0022] Figure 2 This is a schematic diagram of the automatic tensioning control valve structure of the scraper conveyor according to an embodiment of this application;

[0023] Figure 3This is a schematic diagram of fault location coordinates according to an embodiment of this application;

[0024] Figure 4 This is a structural schematic diagram of a fault diagnosis system for an automatic tension control valve of a scraper conveyor according to an embodiment of this application. Detailed Implementation

[0025] This application provides a fault diagnosis method and system for an automatic tension control valve of a scraper conveyor. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0026] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of a fault diagnosis method for an automatic tension control valve of a scraper conveyor in this application includes:

[0027] Step S1: Collect pressure and flow data of the main channel and auxiliary channel of the automatic tensioning control valve through the sensor array, and extract the frequency domain features of the pressure and flow data using signal processing methods to obtain the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel.

[0028] In step S1, obtaining the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel includes: preprocessing the pressure data and flow data to obtain preprocessed time series data; processing the time series data using the Fourier transform algorithm to extract the frequency domain features of the time series data and generate the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel; and normalizing the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel to obtain standardized pressure spectrum of the main channel and flow spectrum of the auxiliary channel.

[0029] Specifically, in the hydraulic drive system of a scraper conveyor, the stable operation of the automatic tension control valve is crucial to ensuring constant tension in the conveyor chain and equipment safety. Internally, it employs a dual-channel design with main and auxiliary channels. The main channel handles the primary pressure load and flow transmission, while the auxiliary channel is responsible for dynamic compensation and fine-tuning. However, internal faults such as valve core jamming and seal wear can disrupt the design coupling between the two channels, leading to abnormal pressure and flow interactions. These fault characteristics typically manifest initially as subtle distortions in the time-domain signal, masked by environmental noise. To achieve accurate early fault diagnosis, the key lies in extracting frequency-domain features from the raw sensor data that characterize the dynamic coupling state between the channels, providing a high signal-to-noise ratio signal foundation for subsequent anomaly identification and fault location.

[0030] In specific implementation, such as Figure 2 The diagram shows the structure of the automatic tensioning control valve for a scraper conveyor. First, a sensor array is deployed and data is acquired on the automatic tensioning control valve. The main flow channel system, represented by a solid rectangular channel, bears the main pressure load and flow transmission functions. High-frequency dynamic pressure sensors P1 and P2 are deployed at the key monitoring sections at the inlet and outlet of the main flow channel to capture pressure data in real time caused by valve core misalignment, internal leakage, or fluid pulsation. The auxiliary flow channel system, represented by a dashed rectangular channel, is responsible for dynamic compensation and fine adjustment. High-precision flow sensors Q1 and Q2 are deployed in the adjustment loop of the auxiliary flow channel to monitor real-time changes in the compensation flow rate. Q1 is located at the inlet of the auxiliary flow channel, and Q2 is located at the outlet, forming a complete flow monitoring network. All sensors are triggered by a unified synchronous clock and synchronously acquire pressure and flow signals at a sampling rate of no less than 1 kHz, forming a strictly time-aligned initial synchronous time series dataset. This high-sampling synchronous acquisition strategy ensures the complete capture of rapid transient processes that may occur in the hydraulic system under dynamic operating conditions, providing data support for subsequent analysis of the instantaneous interaction between the two flow channels. After acquiring the initial time series of pressure and flow data, preprocessing is required to improve data quality and prepare a clean signal source for frequency domain analysis. The preprocessing first employs a median filtering algorithm, which selects the median value through a sliding window to replace the original value, effectively filtering out impulsive spike noise without blurring the signal edges. Subsequently, a weighted moving average algorithm is applied to smooth the signal. By assigning higher weights to recent data points, random high-frequency noise is effectively suppressed while better preserving the trend and main fluctuation characteristics of the signal itself. After this preprocessing process, the denoised and smoothed main flow channel pressure time series data and auxiliary flow channel flow data are obtained.

[0031] Subsequently, the Fast Fourier Transform (FFT) algorithm was used to process the main channel pressure time series. The FFT decomposes the pressure fluctuations in the time domain into the sum of a series of sinusoidal components with different frequencies, phases, and amplitudes, thus obtaining the complex spectrum of the pressure signal. The amplitude spectrum of this complex spectrum is extracted, which is the main channel pressure spectrum. This spectrum intuitively shows the energy distribution of the pressure signal at each frequency component. For example, the periodic micro-vibrations of the valve core may appear as specific low-frequency peaks in the spectrum, while the fundamental frequency pulsations caused by the pump source may appear as another characteristic peak. Similarly, the same FFT processing was performed on the auxiliary channel flow time series data, and its amplitude spectrum was extracted to obtain the auxiliary channel flow spectrum. The auxiliary channel flow spectrum intuitively shows the energy distribution of the flow signal, especially revealing the specific frequency fluctuation patterns caused by changes in the smoothness of the flow path, local blockages, or leaks. To eliminate the impact of differences in sensor range, installation location, and absolute signal strength on subsequent correlation analysis, the obtained pressure and flow spectra need to be normalized. Specifically, a maximum-minimum normalization method is used, dividing the amplitude value of each frequency point in each spectrum by its own maximum amplitude value, so that all amplitude values ​​are linearly mapped to the [0, 1] interval. After this processing, standardized main channel pressure spectra and standardized auxiliary channel flow spectra are obtained. The normalized spectra unify the data scale and dimensions, enabling spectral data from different physical quantities and monitoring points to be compared and correlated on a fair and consistent benchmark. These standardized main channel pressure spectra and auxiliary channel flow spectra, as core features characterizing the frequency domain state of the dual-channel system, are used for the next step of parameter coupling anomaly diagnosis. They reveal the energy distribution pattern of the system at different frequencies in a quantitative form and are the cornerstone for detecting whether the coordination between channels has been disrupted.

[0032] Step S2: Calculate the cross-correlation coefficient based on the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel, and determine whether the cross-correlation coefficient is lower than the preset threshold. If so, it is determined that there is an abnormal parameter coupling between the main channel and the auxiliary channel, and an abnormal coupling index is generated.

[0033] In step S2, generating an abnormal coupling index includes: calculating the cross-correlation coefficient between the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel in the frequency domain; obtaining the peak value of the cross-correlation coefficient and comparing the peak value with a preset threshold. If the peak value is lower than the preset threshold, it is determined that there is an abnormal parameter coupling between the main channel and the auxiliary channel, and an abnormal coupling index is generated. The abnormal coupling index includes the peak value of the cross-correlation coefficient and the corresponding frequency point.

[0034] Specifically, in the coordinated operation of the dual-channel automatic tensioning control valve of the scraper conveyor, there is a strong coupling relationship between the pressure dynamics of the main channel and the flow response of the auxiliary channel, which is the physical basis for the valve to achieve precise pressure regulation and flow compensation. However, early faults such as valve core jamming and seal wear can disrupt this inherent coordination, causing pressure changes to fail to effectively drive the expected flow response, or flow fluctuations to fail to smooth pressure pulsations. Since this detuning effect is weak in the early stages of the fault and easily masked by random noise, it needs to be accurately and quantitatively detected through frequency domain correlation analysis. The core of this analysis lies in calculating the cross-correlation coefficient that characterizes the strength of the dual-channel coordination and determining and quantifying coupling anomalies accordingly. In practical implementation, a standardized pressure spectrum of the main channel is used. Flow spectrum of auxiliary channel As input, where f is a discrete frequency point, and to quantify the morphological similarity between the two in the frequency domain, convolution operation is used to calculate their cross-correlation coefficient sequence. The calculation formula is: ,in, This is the frequency lag. This represents the new sequence obtained by shifting the flow spectrum sequence Q by k units along the frequency axis. Essentially, this operation involves shifting the pressure spectrum sequence Q by k units for each possible hysteresis. Compared with the shifted flow spectrum sequence By performing point-by-point corresponding multiplication and summing the products at all frequency points, a numerical value representing the correlation between the two at that specific offset is obtained. By iterating through all k values ​​within a preset range, typically covering the range from the maximum possible negative frequency offset to the maximum possible positive frequency offset, a complete sequence of cross-correlation coefficients can be obtained. Finally, the maximum value, i.e. the peak value of the cross-correlation coefficient, is extracted from the sequence. The k value corresponding to this peak value is the frequency offset that makes the two spectra reach the optimal alignment state. The magnitude of the peak value directly quantifies the coupling strength of the dual-channel signals in the frequency domain.

[0035] After obtaining the peak value of the cross-correlation coefficient, it needs to be compared with a preset threshold. This threshold is established based on the statistical calculation of a large number of peak cross-correlation coefficient samples under long-term historical normal operating conditions of the control valve. For example, it can be set as a certain percentage of the average value of historical peak values, such as 85%, or determined by subtracting a certain number of standard deviations from the mean. This constructs a statistical boundary characterizing the "lower limit of normal coupling strength". If the calculated current peak value is lower than the preset threshold, it indicates that the synchronization and coordination between the pressure dynamics of the main flow channel and the flow response of the auxiliary flow channel in the frequency domain has been significantly weakened and has deviated from the historical normal range. This state is defined as parameter coupling anomaly, which essentially reveals that the inherent hydraulic-mechanical linkage relationship between the two flow channels has become detuned or decoupled. Once an anomaly is identified, a structured anomaly coupling index is generated. This index aims to quantitatively record the core characteristics of this anomaly event and mainly includes two key data: first, the peak value of the cross-correlation coefficient that triggered the judgment, whose magnitude directly quantifies the severity of the coupling failure; and second, the frequency hysteresis k corresponding to the peak value, which indicates at which specific frequency offset the maximum residual correlation is reached, thus implying the frequency band information of the fault characteristic frequency or system dynamic mismatch that may be related to it. This anomaly coupling index provides preliminary and quantifiable feature inputs for subsequent steps of fault mode classification and refined analysis.

[0036] By transforming the time-domain fluctuation relationship, which is difficult to judge intuitively, into a correlation index that can be accurately calculated and statistically tested in the frequency domain, the transformation of the internal collaborative working state of the dual-channel system from subjective experience judgment to objective quantitative diagnosis has been realized. The generated abnormal coupling index not only provides quantitative evidence for the existence of faults, but also points the way to the initial location of the fault root cause through intensity and frequency information.

[0037] Step S3: Combine the environmental interference data of the scraper conveyor to analyze the abnormal coupling index, obtain the initial distribution range of nonlinear flow characteristics, and correct the initial distribution range by integrating the medium characteristic data. Determine whether the corrected distribution range deviates from the normal reference. If so, output the correction deviation value.

[0038] Step S3, which obtains the initial distribution range of nonlinear flow characteristics, includes: acquiring environmental disturbance data of the scraper conveyor; fusing the abnormal coupling index and environmental disturbance data to obtain a fused feature vector; using a support vector machine algorithm to classify and train the fused feature vector to obtain a classification model; judging the fused feature vector based on the trained classification model, and outputting the category label of the current system state and the corresponding classification confidence; extracting potential patterns reflecting the dynamic changes in pressure differences between the main channel and auxiliary channels from the discrimination results of the classification model; statistically analyzing the distribution patterns of the corresponding feature values ​​under historical normal and abnormal operating conditions based on the potential patterns; determining the initial distribution range of nonlinear flow characteristics in the potential patterns; the initial distribution range includes the upper and lower limits of the feature values ​​and the probability density within the entire numerical range; and performing statistical analysis on the distribution range to obtain the statistical parameters of the distribution range.

[0039] Specifically, in the actual operating environment of a scraper conveyor, the working state of the automatic tensioning control valve is affected not only by its internal mechanical coupling relationship but also by external environmental factors, especially temperature fluctuations. Temperature changes cause alterations in the viscosity of the hydraulic medium and subtle thermal deformations in the valve body structure. These effects are superimposed on the actual fault characteristics, manifesting as complex nonlinear changes in pressure and flow signals. To accurately distinguish between actual fault signals and environmental interference, and to extract the nonlinear characteristics characterizing the essential flow state within the valve, it is necessary to fuse and analyze abnormal coupling indicators reflecting the internal coupling state with external environmental interference data. In practice, temperature time-series data of the scraper conveyor's operating environment is first collected in real time using temperature sensors installed near the valve body or flow channel. This data serves as a key source of environmental interference, reflecting the comprehensive thermal load changes in the valve's environment. Next, the abnormal coupling index generated in the previous step is fused with the temperature disturbance data. The fusion process adopts a weighted splicing method. For example, in order to highlight the dominance of internal coupling anomaly, the abnormal coupling index vector is given a weight of 0.7 and the normalized temperature fluctuation data vector is given a weight of 0.3. The two weighted vectors are connected into a high-dimensional fusion feature vector. This vector comprehensively represents the quantitative information of the coupling state of the two flow channels inside the valve under a specific ambient temperature background. Using fused feature vectors as training samples, a support vector machine (SVM) algorithm is employed for supervised learning to construct a classification model capable of distinguishing different states of automatic tension control valves. The SVM uses kernel function techniques to handle nonlinear separability issues in the data; in this implementation, a radial basis function (RBF) kernel is preferred. The fused feature vectors are mapped to a high-dimensional feature space, where the optimal classification hyperplane is found. The model is trained using historical data containing fused feature vectors under different temperatures and fault degrees, along with their labeled state tags. The trained model can output a category label for the state of the automatic tension control valve, such as "normal," "valve core jamming tendency," or "seal wear tendency," along with its corresponding classification confidence score, based on newly input real-time fused feature vectors. This confidence score reflects the model's degree of certainty in determining the current state.

[0040] The potential patterns reflecting the dynamic changes in pressure differences are extracted from the real-time discrimination results of the classification model. The potential patterns refer to the typical change patterns that are summarized from the dynamic data of pressure differences and correspond to specific fault precursors or stable operating conditions, such as: progressive jamming precursor patterns, micro-leakage oscillation patterns, etc. These defined potential patterns provide a structured analysis object and benchmark for subsequent accurate quantification of fault characteristics, statistical analysis of their distribution patterns and accurate diagnosis. The determination of potential patterns will be explained in detail later.

[0041] For each defined potential pattern, its corresponding nonlinear flow characteristics need to be quantified. Nonlinear flow characteristics refer to the physical phenomenon in a two-channel system where the dynamic response relationship between pressure and flow deviates from the linear proportional law of its healthy state, caused by internal mechanical abnormalities such as valve core jamming or seal wear. This deviation is specifically and repeatedly reflected in the pressure difference signal between the main channel and the auxiliary channel, forming a specific dynamic change pattern, i.e., a potential pattern. Therefore, quantifying nonlinear flow characteristics is essentially a process of mathematically extracting the key signal attributes characterizing the potential pattern. Taking the "progressive jamming precursor pattern" as an example, this pattern is characterized by a slow increase in the mean pressure difference signal accompanied by an increase in low-frequency energy. To quantify this phenomenon, an attribute that can capture its core dynamics—the mean pressure difference window—can be extracted from all historical data segments clustered into this pattern. Specifically, for each data window belonging to this pattern, the arithmetic mean of all pressure difference sampling points within it is calculated. This calculated mean is the specific quantification value of the nonlinear flow characteristic of a gradual increase in pressure under this pattern. Similarly, for other potential modes, such as the "high-frequency micro-leakage oscillation mode," its nonlinear flow characteristics can be quantified as the "pressure difference signal standard deviation" or the "energy proportion of a specific high-frequency band" within the data window under that mode. In this way, each potential mode is associated with one or more specific, computable feature values, which serve as the direct data objects for subsequent in-depth probability distribution modeling and statistical parameter analysis.

[0042] Based on latent patterns, statistical analysis is performed on the distribution patterns of corresponding feature values ​​under historical normal and abnormal operating conditions. The aim is to establish a quantitative statistical benchmark for each identified fault precursor pattern. This process involves two levels: First, a set of feature values ​​collected under operating conditions marked as "normal" is gathered. By analyzing the distribution of specific feature values, such as the mean of the pressure difference window, in a large amount of historical normal data, its normal statistical range, such as mean and variance, can be established. When a new data segment is clustered into a certain latent pattern by a clustering algorithm, it essentially means that its feature values ​​have significantly deviated from this normal reference range, thus being judged as "abnormal." Then, the second level proceeds by collecting the same feature values ​​identified under the specific latent pattern, i.e., abnormal operating conditions, forming an abnormal feature value set. For this abnormal set, non-parametric methods such as kernel density estimation are used to fit its continuous probability density function, thereby bypassing prior assumptions about the distribution form. Based on this estimated probability density function, its cumulative distribution is calculated to determine a numerical boundary covering a preset high probability interval, such as 95%. This boundary value is the lower and upper limits of the initial distribution range. For example, for a "progressive lag precursor mode," the initial distribution range of the mean pressure difference window might be quantified as a lower limit of -2.1 and an upper limit of 3.5. Finally, statistical analysis is performed on this distribution range defined by the anomalous data to obtain its statistical parameters. This mainly involves calculating the expected value and standard deviation of the anomalous characteristic value within this range. The expected value reflects the average or central trend of the characteristic value under this anomalous mode, while the standard deviation quantifies the degree of fluctuation around this central trend. These statistical parameters, along with the upper and lower limits and the probability density function, constitute a complete statistical description of the nonlinear flow characteristics of this potential mode, providing an initial, quantitative benchmark model for subsequent integration with other data for dynamic correction and refinement.

[0043] The process involves extracting potential patterns reflecting the dynamic changes in pressure differences between the main flow channel and auxiliary flow channels. This includes: parsing feature vector sequences representing the trend of pressure difference changes from the category labels and classification confidence scores output by the classification model; performing a sliding window analysis on the feature vector sequences over time to extract statistical features, frequency domain features, and time domain correlation features of the data within the window, thus constructing a multidimensional potential pattern feature set; and identifying recurring pressure difference change patterns through cluster analysis based on the potential pattern feature set, defining each pattern and its corresponding feature value range as a potential pattern.

[0044] Specifically, during real-time operation, the classification model outputs a system state category label and its corresponding classification confidence score for each sampling moment. This time-series discrimination result is aligned and fused with the real-time pressure difference values ​​between the main flow channel and auxiliary flow channels, which are simultaneously collected and calculated, to construct a multi-dimensional feature vector sequence. Each vector in this sequence contains the instantaneous value of the pressure difference, the quantized encoding of the state category, and the confidence score, thus integrating the changes in physical quantities with the preliminary judgment of intelligent diagnosis into a unified time-series data. Subsequently, a sliding window technique is used to segment the time-series data, and multi-dimensional features are extracted within each time window. These include statistical features such as mean and variance; frequency domain features such as the main frequency amplitude obtained through FFT and the energy proportion of a specific frequency band; and time-domain correlation features such as the autocorrelation coefficient. All features extracted from each window constitute a multi-dimensional latent pattern feature set, with each feature point concisely representing the core dynamic form of the original data within the corresponding time period.

[0045] Based on the multidimensional potential pattern feature set, cluster analysis is used to automatically identify the recurring typical patterns, which are closely related to specific valve mechanical conditions or early signs of failure. Specifically, clustering algorithms such as K-means or DBSCAN can be used to divide feature points into several clusters based on their proximity and density distribution in multidimensional space. The feature points within each cluster exhibit high similarity, collectively corresponding to a specific dynamic pattern of pressure difference that has repeatedly appeared in historical data. For example, the algorithm might identify a pattern characterized by "a gradual increase in the mean pressure difference, a gradual increase in the standard deviation, and a continuously rising proportion of low-frequency energy." This pattern repeatedly appears in historical data during the early stages of progressive valve core jamming and can be defined as a "prodromal jamming pattern." Another identified pattern might manifest as "high-frequency, small-amplitude oscillations in pressure difference, periodic shifts in the mean, and weak autocorrelation." This pattern is often associated with intermittent minor leaks caused by seal wear and can be defined as a "minor leak oscillation pattern." Furthermore, there might be a pattern of "slow recovery after a sudden step change in pressure difference," which could correspond to a transient event of a foreign object passing through the valve cavity. Each clearly defined potential pattern corresponds to a specific cluster in the clustering feature space and possesses a typical numerical range for its multidimensional feature values. This process ultimately reduces the complex and dynamic behaviors implicit in the data stream to a series of limited, clearly describable, and indexable typical patterns, providing a structured knowledge base for subsequent accurate fault location and condition assessment.

[0046] In step S3, the initial distribution range is corrected by fusing medium characteristic data. This includes: collecting viscosity data of the hydraulic working medium flowing through the main channel and auxiliary channel of the automatic tension control valve as medium characteristic data; fusing the medium characteristic data based on the distribution range of nonlinear flow characteristics using a Kalman filter algorithm to obtain the corrected distribution range; comparing the corrected distribution range with a preset normal benchmark and calculating the deviation between the two; if the deviation exceeds a preset threshold, it is determined that the corrected distribution range deviates from the normal benchmark, and a correction deviation value is generated; and the correction deviation value is normalized to obtain a standardized correction deviation value.

[0047] Specifically, in the actual operation of scraper conveyors, the viscosity of the hydraulic working medium in the automatic tensioning control valve fluctuates dynamically with changes in ambient temperature, load, and the aging of the medium itself. This viscosity change directly modulates the pressure and flow relationship within the flow channel, thus interfering with the judgment of purely mechanical fault characteristics. To isolate the influence of medium characteristics and obtain features that only reflect the mechanical state of the valve body, it is necessary to dynamically correct the initial distribution range of nonlinear flow characteristics obtained in the aforementioned steps. In practice, firstly, an online viscosity sensor installed in the flow channel collects the current viscosity value of the hydraulic working medium flowing through the valve in real time, serving as key medium characteristic data. This data reflects the key physical properties affecting the flow characteristics at the current moment. Next, the initial distribution range of nonlinear flow characteristics and its statistical parameters, such as mean and variance, are used as the state vector for modeling the Kalman filter algorithm. The Kalman filter, as an optimal recursive estimation algorithm, is used here to achieve data fusion between the state (reflecting the mechanical state distribution range) and the observed values ​​(the medium viscosity data). The algorithm runs in two steps: the prediction step relies on a system dynamics model describing the dynamic relationship between viscosity and flow characteristics. This model is usually based on fluid mechanics principles and valve characteristics, and can be expressed as a discrete-time state-space equation: , where the state vector Parameters representing the range of nonlinear flow characteristics at the current moment, such as mean and variance, It is the state estimate from the previous moment, the control input. For the currently measured medium viscosity data, it is input through a matrix. Directly affecting state prediction, the state transition matrix A encodes the inertia or memory effect of the system state, and process noise. This characterizes the uncertainty of the model. The prediction step uses this model, combining the optimal state estimate from the previous time step with the current viscosity input, to calculate the prior predicted value of the current state and its prediction error covariance matrix. The output prior state estimate of this step is the distribution range parameter predicted based on historical states and current medium conditions, without correction by the current observation and subsequent update steps. It forms the basis for data fusion and state updates. The update step then introduces real-time measured medium viscosity data, calculates the Kalman gain to balance the reliability of the predicted value with the new observation information, and thus corrects the predicted distribution range, outputting the corrected distribution range. For example, when the medium viscosity increases significantly, the algorithm will correspondingly "shrink" the characteristic distribution range to offset the potential pressure increase, similar to slight blockage, that might be caused by the increased viscosity itself.

[0048] Subsequently, the corrected distribution range is compared with a preset normal baseline. This normal baseline is established by statistically analyzing the characteristic distribution range exhibited by the valve during long-term historical normal operation under different media viscosity conditions; for example, it may be a mean range of characteristic values ​​[0.1, 0.9]. A quantified deviation value is obtained by calculating the central tendency of the corrected distribution range, such as the Euclidean distance between the mean and the center of the normal baseline range, or by calculating the statistical distance between their probability distributions, such as the KL divergence. This deviation value characterizes the degree of deviation between the corrected characteristic and the historical normal state. The calculated deviation value is compared with a preset deviation threshold, which is determined based on the analysis of historical normal fluctuation ranges. For example, it can be set as the 95th percentile of historical normal deviation values. If the current deviation value exceeds this threshold, it is judged as a statistically significant deviation, indicating an anomaly that cannot be explained by changes in the medium. At this point, a corrected deviation value is generated, which is the quantified value of the portion exceeding the threshold. It directly reflects the intensity of the symptoms caused by mechanical failure. For example, if the threshold is 0.15 and the deviation value is 0.22, then a corrected deviation value of 0.22 is generated. Finally, the correction deviation value is normalized, for example, by linearly scaling it to the [0, 1] interval to obtain a standardized correction deviation value. This standardized value eliminates the dimensional differences under different fault modes or operating conditions, providing a comparable input for subsequent similarity calculation with a preset fault mode of a unified scale. The whole process effectively improves the robustness of fault diagnosis features to changes in operating conditions by dynamically fusing media information, allowing the diagnostic focus to be on the health status of the mechanical components themselves.

[0049] Step S4: Calculate the similarity score between the correction deviation value and the preset fault mode. If the similarity score is higher than the preset threshold, determine the corresponding fault type. Based on the similarity score, generate fault location coordinates through optimization and mapping processing.

[0050] In step S4, determining the corresponding fault type includes: acquiring standardized feature data of a preset fault mode, the standardized feature data including the pressure-flow deviation feature spectrum corresponding to the seal wear mode; calculating the similarity score between the correction deviation value and the feature data of the preset fault mode; determining whether the similarity score is higher than a preset threshold, and if the similarity score is higher than the preset threshold, then the corresponding preset fault mode is determined as the current fault type.

[0051] Specifically, to achieve accurate identification of fault types in the automatic tensioning control valve of the scraper conveyor, the core lies in quantifying and comparing the standardized correction deviation value with the standardized feature data of the preset fault mode, thereby completing the mapping from continuous signals to a clear fault category. In practice, the first step is to acquire the standardized feature data of the preset fault mode. This data is constructed through in-depth mining and analysis of a historical fault case database. For each typical fault mode, such as the "seal wear mode," dynamic signal features that essentially reflect the fault are extracted from historical cases. These features include, but are not limited to, specific pressure drop gradients, characteristic frequency components of pressure pulsations, steady-state offset of auxiliary flow channel compensation flow, and its fluctuation spectrum characteristics. These multi-dimensional features are normalized to eliminate dimensional differences in absolute values, and then combined into a standardized, multi-dimensional pressure-flow deviation feature spectrum. This feature spectrum constitutes the "baseline coordinates" of the fault mode in the feature space. In real-time diagnosis, the correction deviation value obtained after media characteristic correction in the previous steps is used to construct a corresponding real-time feature vector based on the dimensional structure of the feature spectrum. Each element in this vector corresponds to the quantified representation of the correction deviation value in a certain feature dimension. Next, the similarity score between the correction deviation value and the feature data of the preset fault mode is calculated. This calculation usually uses the cosine similarity algorithm, which calculates the cosine value of the angle between two feature vectors, namely the real-time feature vector and the standardized feature spectrum vector. Its value range is [0, 1]. The closer the score is to 1, the more consistent the directions of the two in the multi-dimensional feature space are, that is, the more similar the current system state is to the morphological features of the target fault mode.

[0052] After calculating the similarity score, it is determined whether the similarity score is higher than a preset threshold. This preset threshold is not an empirical constant but is dynamically determined based on statistical analysis of historical data. Specifically, for each preset fault mode, the similarity score between the real-time feature vectors of all confirmed historical cases and the standard feature spectrum of that mode is calculated, forming a historical score sample set. Statistical analysis is performed on this sample set, and a high percentile of its distribution, such as the 95th percentile or "mean + 2 standard deviations," is taken as the preset threshold for that fault mode. This setting ensures that the threshold can cover the vast majority of real fault scenarios while effectively resisting accidental high scores caused by random noise or unmodeled interference. If the currently calculated similarity score is higher than the preset threshold, it statistically indicates that the feature matching degree between the current system state and the target preset fault mode has reached a significant level, thus reliably identifying the corresponding preset fault mode, such as "seal wear mode," as the current fault type. Through this series of steps based on quantitative features and statistical inference, a precise and reliable mapping from continuous monitoring data to discrete, clearly defined fault types is achieved.

[0053] In step S4, fault location coordinates are generated through optimization and mapping, including: using a genetic algorithm to iteratively optimize the similarity score, with the goal of maximizing the similarity score, and outputting an optimized fault feature parameter set; extracting multidimensional fault feature vectors from the optimized fault feature parameter set, and using principal component analysis to reduce the dimensionality of the multidimensional fault feature vectors to obtain dimensionality-reduced feature vectors; and mapping the dimensionality-reduced feature vectors to a predefined fault location coordinate system associated with the physical structure of the automatic tension control valve to generate fault location coordinates.

[0054] Specifically, in the fault diagnosis of the automatic tensioning control valve of the scraper conveyor, in order to transform the diagnostic conclusion from an abstract similarity score into coordinates that can be accurately located in the three-dimensional space of the valve body, a series of mathematical processes involving optimization, dimensionality reduction, and physical mapping need to be performed. The core of this process lies in constructing a robust mapping from a high-dimensional fault feature space to a low-dimensional physical location space. In practice, the similarity score is first iteratively optimized using a genetic algorithm. Genetic algorithms are a global optimization method that simulates the natural selection and genetic mechanisms of organisms. In this scenario, the optimization objective is to maximize the similarity score. The initial population of the algorithm consists of a series of candidate fault feature parameter sets. Each individual, i.e., a parameter set, represents a possible interpretation of the current fault state. By defining a fitness function with similarity score as the evaluation criterion, the algorithm performs selection, crossover, and mutation operations during the iteration process: the selection operation retains individuals with high fitness; the crossover operation swaps some parameters of excellent individuals to generate new individuals to explore potential better solutions; the mutation operation randomly perturbs some parameters of individuals to maintain population diversity and avoid premature convergence to local optima. After multiple generations of evolution, the algorithm finally converges and outputs the optimized fault feature parameter set. Mathematically, this parameter set represents the feature combination that achieves the highest matching degree between the system state and the target fault mode.

[0055] Subsequently, from the optimized fault feature parameter set, all key parameters are extracted to form a multidimensional fault feature vector. This vector is typically high-dimensional, and there may be correlations or information redundancy between dimensions. To extract the most essential fault information and simplify subsequent processing, principal component analysis (PCA) is used to reduce the dimensionality of this multidimensional vector. The PCA algorithm calculates the covariance matrix of the vector and performs eigenvalue decomposition to identify a new set of orthogonal principal component directions, which are ordered according to the magnitude of the variance explained in the original data. By selecting the top k principal components whose cumulative variance contribution rate exceeds a preset threshold, such as 95%, the original high-dimensional vector is projected onto the low-dimensional subspace formed by these k principal components, thus obtaining the dimensionality-reduced feature vector. This operation significantly reduces data dimensionality and noise interference while retaining most of the original information. Finally, the dimensionality-reduced feature vectors are mapped to a predefined fault location coordinate system precisely associated with the physical structure of the automatic tension control valve to generate the final fault location coordinates. This coordinate system is pre-established based on the valve's three-dimensional digital model, and its origin and axial direction are strictly aligned with key geometric references of the actual valve body, such as the valve core centerline and end face. A pre-defined, calibrated mapping function, such as a regression model or lookup table trained on a large number of historical fault cases and corresponding physical location data, converts each numerical component of the dimensionality-reduced feature vector into a coordinate value in the corresponding direction within this physical coordinate system. Figure 3As shown, this is a schematic diagram of fault location coordinates. The coordinates of the location points (such as X=0.15, Y=0.20, Z=0.35) are marked on the three-dimensional structure of the virtual valve, which helps to understand how to accurately mark the fault point in three-dimensional space. The fault location coordinates generated directly point to the specific location where the fault is most likely to occur in the three-dimensional space inside the valve, providing accurate spatial guidance for subsequent virtual simulation verification and precise maintenance intervention.

[0056] Step S5: Map the fault location coordinates to the virtual model of the automatic tension control valve, and verify the stability of the fault location coordinates through dynamic simulation. If the stability index meets the preset conditions, generate a real-time warning signal containing the fault type and fault location coordinates.

[0057] In step S5, the stability of the fault location coordinates is verified through dynamic simulation. This includes: constructing a three-dimensional virtual model containing the main flow channel, auxiliary flow channels, and their connections based on the geometric structure and physical parameters of the automatic tension control valve, and mapping the fault location coordinates to the corresponding internal position of the flow channel in the virtual model; using real-time collected pressure data, flow data, and medium viscosity data as boundary conditions, simulating the dynamic fluid interaction process of the main flow channel and auxiliary flow channels using a dynamic working condition simulation algorithm based on computational fluid dynamics principles, and obtaining simulation results; extracting the flow field parameter sequence related to the fault location coordinates from the simulation results, calculating the statistical variance of the flow field parameter sequence, and defining the variance value as the stability index; comparing the stability index with the preset stability conditions, and if the stability index meets the preset stability conditions, then the fault location coordinates are determined to be stable.

[0058] Specifically, to ensure the physical rationality and diagnostic reliability of the fault location coordinates, after generating the coordinates, they need to be placed in a high-fidelity digital virtual environment for dynamic stability verification under operating conditions. The core of this process lies in using computational fluid dynamics simulation to simulate the interaction between the flow field and the suspected fault point under real operating conditions, thereby assessing whether the location coordinates characterize a stable and significant anomaly source. In specific implementation, firstly, a three-dimensional virtual model is constructed based on the precise three-dimensional geometry and physical parameters of the automatic tension control valve. The three-dimensional geometry includes the precise dimensions, shape, and connection relationships of the main flow channel and auxiliary flow channels, as well as models of key internal components such as the valve core and seals. Physical parameters include material properties and surface roughness. Subsequently, the fault location coordinates generated in the previous steps are precisely mapped to the corresponding three-dimensional spatial position inside the virtual model according to the preset mapping relationship between their coordinate values ​​and the virtual model coordinate system. This position represents the suspected fault point identified by the diagnostic system, for example, it may be located at a specific circumferential and axial position of a sealing ring. Next, using the virtual model as a carrier, a dynamic operating condition simulation algorithm based on computational fluid dynamics principles is used to simulate the transient flow inside the valve. The simulation boundary conditions directly use real-time synchronously acquired pressure data as pressure inlet or outlet conditions, flow rate data as flow rate boundary conditions, and medium viscosity data to define fluid properties. The simulation algorithm solves the mass conservation (continuity equation) and momentum conservation (Navier-Stokes equations) equations describing fluid motion, iteratively calculating the velocity field, pressure field, and other flow field information that change with time within the model under given boundary conditions. This process simulates the dynamic fluid interaction behavior of the main channel and auxiliary channels under actual working conditions and outputs a simulation result dataset containing complete spatiotemporal information. The simulation result dataset contains complete flow field state data of all computational grid cells or nodes in the virtual model at each simulation time step, specifically including three-dimensional velocity vectors, pressure scalar values, turbulence parameters such as turbulent kinetic energy and dissipation rate at each point in space, and key parameters derived from these basic field quantities for evaluating stability, such as the pressure fluctuation sequence at monitoring points. This dataset records the full spatiotemporal evolution information of the dynamic interaction process, providing a data foundation for subsequently extracting parameter sequences at specific locations to calculate the stability index.

[0059] Then, from the simulation result dataset, a sequence of local flow field parameters associated with the fault location coordinates is specifically extracted. Specifically, in the virtual model, a small spherical or cubic spatial region is defined as a "monitoring probe" centered on the fault location coordinates. A key flow field parameter of this "probe" region is extracted over the entire simulation time history, such as the sequence of transient pressure value, velocity vector amplitude, or turbulent kinetic energy changes over time. The statistical variance of this time series is calculated, and the variance value is defined as the stability index. This index quantifies the degree of fluctuation of the flow state near the fault point in the simulated dynamic flow field: the larger the variance, the more severe the flow state at that point is affected by global flow disturbances, and the worse the stability of the location coordinates; the smaller the variance, the more stable the flow state at that point is, and the location coordinates may correspond to a stable physical anomaly, such as a fixed leak point or jamming location. Finally, the calculated stability index is compared with a preset stability condition, which is determined based on statistical analysis of historical data: by performing similar simulations of the valve under a large number of historical normal operating conditions and known slight disturbance conditions, the stability index of any location or specific type of fault point in its virtual model is calculated, its distribution is statistically analyzed, and a threshold is set, for example, the 95th percentile of the historical stability index distribution. If the stability index of the current fault point is lower than this preset threshold, it is determined that the preset stability condition is met, and the fault location coordinate is determined to be stable. This indicates that the anomaly shown by the coordinate point in the dynamic simulation is continuous and significant, rather than caused by random fluctuations or simulation errors.

[0060] Conversely, if the current stability index is higher than this preset threshold, it indicates that the stability index of the fault location coordinates does not meet the preset conditions. In this case, a model update and self-learning process is executed, including: recording the current diagnostic event data package that failed stability verification, containing the current fault type, fault location coordinates, the real-time sensor data sequence that triggered the diagnosis, the calculated correction deviation value, the similarity score, and the stability index; retrieving historical event data from the historical diagnostic event database that has the same fault type or similar operating conditions as the current diagnostic event data package, including verified stable events and events that failed verification; constructing a self-learning feature set for the retrieved current and historical events, including: the spatial encoding of the fault location coordinates, the frequency domain feature vector representing the pressure-flow coupling state, and environmental and medium parameters; analyzing the self-learning feature set using an unsupervised clustering algorithm, dividing the events into different clusters based on feature similarity, identifying feature combination types, including those leading to stable diagnostic coordinates and those prone to coordinate instability or verification failure; generating or adjusting key system parameters based on the clustering analysis results; and repeating steps S3-S5 based on the adjusted key system parameters until the stability index meets the preset conditions. Based on this stability verification conclusion, combined with the identified fault types and verified fault location coordinates, a real-time early warning signal containing this core information is generated and sent, thereby completing the entire process from intelligent diagnosis to reliable early warning.

[0061] Specifically, if the calculated stability index in the virtual simulation verification is higher than a preset threshold, the stability verification of the fault location coordinates is deemed to have failed. In this case, an adaptive update and self-learning process for the diagnostic model will be initiated. This process first stores the complete diagnostic event that failed verification, including the fault type, location coordinates, the original sensor data sequence that triggered the diagnosis (e.g., pressure, flow rate, medium viscosity), calculation process data (e.g., correction deviation value), similarity score, and the final stability index, as a structured "unverified event data package". Simultaneously, it retrieves historical cases similar to the current event in terms of fault type or key operating parameters from the historical diagnostic event database. These cases also include both verified stable events and unverified events. Then, a multi-dimensional feature set for self-learning is constructed. This feature set is extracted from the data packages of the current event and the retrieved historical events, covering the spatial encoding of the fault coordinates, the core frequency domain feature vector reflecting the dynamic coupling relationship of pressure and flow, and the corresponding ambient temperature and medium viscosity parameters.

[0062] Unsupervised clustering algorithms are used to analyze this multidimensional feature set, automatically dividing events into different clusters based on feature similarity. This clustering process can reveal implicit pattern associations. For example, it may identify a cluster mainly composed of "coordinate stable and successful verification" events, characterized by a specific parameter combination; at the same time, it may also identify another cluster composed of "coordinate unstable or failed verification" events, characterized by a different parameter pattern. Based on the patterns revealed by clustering analysis, targeted optimization and adjustment of internal key parameters are performed, mainly including: updating the standardized feature data of the preset fault modes to make them closer to the commonalities of stable fault cases; correcting the empirical parameters or boundary conditions in the system dynamics model on which the dynamic operating condition simulation algorithm depends; and recalibrating the preset threshold for stability judgment to make it more consistent with the statistical distribution of actual operating data. After completing parameter optimization, the system uses the updated model parameters to re-execute the diagnostic process from feature analysis in step S3, fault matching and location in step S4, and stability verification in step S5 on the current or new monitoring data. This iterative process can be repeated until the diagnostic results of the new data can pass stability verification, thereby ensuring that the final output warning signal is based on a highly reliable diagnostic model that has been continuously learned and optimized. Only after the fault location coordinates pass stability verification is a real-time warning signal containing the verified fault type and precise location coordinates generated and sent, thus completing the closed loop from intelligent diagnosis to reliable decision-making.

[0063] In summary, this application provides a full-process, high-precision intelligent fault diagnosis method for automatic tension control valves of scraper conveyors. This method deploys a high-sampling sensor array to synchronously acquire pressure and flow signals from both flow channels. After filtering, Fourier transform, and normalization, standardized frequency-domain pressure and flow spectra are obtained, providing a high signal-to-noise ratio characteristic basis for capturing weak coupling anomalies. Utilizing the peak value and statistical threshold of the frequency-domain cross-correlation coefficient, objective and quantitative detection of coupling anomalies in the dual-flow channel parameters is achieved. By fusing temperature interference and medium viscosity data, combined with support vector machine classification and Kalman filter correction, the interference of environmental and medium disturbances on diagnostic features is effectively eliminated, accurately extracting the nonlinear flow characteristics and their statistical distribution reflecting the mechanical state. Furthermore, by matching the correction deviation with the standard fault feature spectrum using cosine similarity, and combining genetic algorithm optimization and PCA dimensionality reduction mapping, the abstract fault matching degree is transformed into precise physical coordinates in the three-dimensional space of the valve body. Finally, by establishing a high-fidelity virtual model and performing dynamic simulation based on computational fluid dynamics, the stability of the located fault coordinates is verified, ensuring the physical rationality and reliability of the diagnostic conclusions. This method realizes a closed-loop diagnosis across the entire chain, from data acquisition, feature extraction, anomaly detection, fault identification, precise positioning to virtual verification. It significantly improves the early detection, accuracy, and adaptability of fault diagnosis in complex hydraulic systems, providing reliable technical support for predictive maintenance of equipment.

[0064] The above describes a fault diagnosis method for an automatic tension control valve of a scraper conveyor according to an embodiment of this application. The following describes a fault diagnosis system for an automatic tension control valve of a scraper conveyor according to an embodiment of this application. Please refer to [link / reference]. Figure 4 One embodiment of the fault diagnosis system for the automatic tensioning control valve of a scraper conveyor in this application includes:

[0065] The data acquisition module is used to collect pressure and flow data of the main channel and auxiliary channel of the automatic tension control valve through a sensor array, and to extract the frequency domain features of the pressure and flow data using signal processing methods to obtain the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel.

[0066] The anomaly analysis module is used to calculate the cross-correlation coefficient based on the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel, and to determine whether the cross-correlation coefficient is lower than the preset threshold. If so, it is determined that there is a parameter coupling anomaly between the main channel and the auxiliary channel, and an anomaly coupling index is generated.

[0067] The correction processing module is used to analyze the abnormal coupling index by combining the environmental interference data of the scraper conveyor, obtain the initial distribution range of nonlinear flow characteristics, and correct the initial distribution range by integrating the medium characteristic data. It then determines whether the corrected distribution range deviates from the normal reference. If so, it outputs the correction deviation value.

[0068] The preliminary positioning module is used to calculate the similarity score between the correction deviation value and the preset fault mode. If the similarity score is higher than the preset threshold, the corresponding fault type is determined. Based on the similarity score, fault positioning coordinates are generated through optimization and mapping processing.

[0069] The stability verification module is used to map the fault location coordinates to the virtual model of the automatic tension control valve and verify the stability of the fault location coordinates through dynamic simulation. If the stability index meets the preset conditions, a real-time warning signal containing the fault type and fault location coordinates is generated.

[0070] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0071] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0072] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A fault diagnosis method for an automatic tension control valve of a scraper conveyor, characterized in that, The method includes: Step S1: Collect pressure and flow data of the main channel and auxiliary channel of the automatic tensioning control valve through a sensor array, and extract the frequency domain features of the pressure and flow data using signal processing methods to obtain the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel. Step S2: Calculate the cross-correlation coefficient based on the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel, and determine whether the cross-correlation coefficient is lower than a preset threshold. If so, determine that there is a parameter coupling anomaly between the main channel and the auxiliary channel, and generate an abnormal coupling index. Step S3: Analyze the abnormal coupling index using environmental interference data from the scraper conveyor to obtain the initial distribution range of the nonlinear flow characteristics. Correct the initial distribution range using fused medium characteristic data. Determine whether the corrected distribution range deviates from the normal baseline. If so, output the correction deviation value. Step S3, obtaining the initial distribution range of the nonlinear flow characteristics, includes: acquiring environmental interference data from the scraper conveyor; fusing the abnormal coupling index and the environmental interference data to obtain a fused feature vector; using a support vector machine algorithm to classify and train the fused feature vector to obtain a classification model; judging the fused feature vector based on the trained classification model, outputting the category label of the current system state and the corresponding classification confidence; extracting potential patterns reflecting the dynamic changes in pressure differences between the main flow channel and auxiliary flow channels from the discrimination results of the classification model; statistically analyzing the distribution patterns of the corresponding feature values ​​under historical normal and abnormal operating conditions based on the potential patterns; determining the initial distribution range of the nonlinear flow characteristics in the potential patterns, where the initial distribution range includes the upper and lower limits of the feature values ​​and the probability density within the entire numerical range; and performing statistical analysis on the initial distribution range to obtain statistical parameters of the distribution range. Step S4: Calculate the similarity score between the correction deviation value and the preset fault mode. If the similarity score is higher than the preset threshold, determine the corresponding fault type. Based on the similarity score, generate fault location coordinates through optimization and mapping processing. Step S5: Map the fault location coordinates to the virtual model of the automatic tension control valve, and verify the stability of the fault location coordinates through dynamic simulation. If the stability index meets the preset conditions, generate a real-time warning signal containing the fault type and the fault location coordinates.

2. The method according to claim 1, characterized in that, Step S1 yields the pressure spectrum of the main flow channel and the flow spectrum of the auxiliary flow channel, including: The pressure data and flow data are preprocessed to obtain preprocessed time series data; the time series data are processed using a Fourier transform algorithm to extract the frequency domain features of the time series data, generating the main channel pressure spectrum and the auxiliary channel flow spectrum; the main channel pressure spectrum and the auxiliary channel flow spectrum are normalized to obtain standardized main channel pressure spectrum and auxiliary channel flow spectrum.

3. The method according to claim 1, characterized in that, Step S2 generates an abnormal coupling index, including: For the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel, calculate the cross-correlation coefficient between them in the frequency domain; obtain the peak value of the cross-correlation coefficient and compare the peak value with a preset threshold. If the peak value is lower than the preset threshold, it is determined that there is a parameter coupling anomaly between the main channel and the auxiliary channel, and an abnormal coupling index is generated. The abnormal coupling index includes the peak value of the cross-correlation coefficient and the corresponding frequency point.

4. The method according to claim 1, characterized in that, Extract potential patterns reflecting the dynamic changes in pressure differences between the main flow channel and auxiliary flow channels, including: From the category labels and classification confidence scores output by the classification model, a feature vector sequence representing the trend of pressure difference changes is parsed out; a sliding window analysis is performed on the feature vector sequence in the time dimension to extract the statistical features, frequency domain features, and time domain correlation features of the data within the window, forming a multidimensional latent pattern feature set; based on the latent pattern feature set, a clustering analysis method is used to identify recurring patterns of pressure difference changes, and each pattern and its corresponding feature value range are defined as a latent pattern.

5. The method according to claim 1, characterized in that, Step S3 involves correcting the initial distribution range using the fused medium characteristic data, including: Viscosity data of the hydraulic working medium flowing through the main channel and auxiliary channel of the automatic tensioning control valve are collected as medium characteristic data; based on the distribution range of the nonlinear flow characteristics, the medium characteristic data are fused using a Kalman filter algorithm to obtain a corrected distribution range; the corrected distribution range is compared with a preset normal benchmark, and the deviation value between the two is calculated. If the deviation value exceeds a preset threshold, the corrected distribution range is determined to deviate from the normal baseline, and a correction deviation value is generated; the correction deviation value is normalized to obtain a standardized correction deviation value.

6. The method according to claim 1, characterized in that, Step S4 determines the corresponding fault type, including: Obtain standardized feature data of a preset fault mode, the standardized feature data including the pressure-flow deviation feature spectrum corresponding to the seal wear mode; calculate the similarity score between the correction deviation value and the feature data of the preset fault mode; determine whether the similarity score is higher than the preset threshold, and if the similarity score is higher than the preset threshold, determine the corresponding preset fault mode as the current fault type.

7. The method according to claim 6, characterized in that, Step S4 generates fault location coordinates through optimization and mapping processes, including: A genetic algorithm is used to iteratively optimize the similarity score, with maximizing the similarity score as the optimization objective, and an optimized fault feature parameter set is output. Multidimensional fault feature vectors are extracted from the optimized fault feature parameter set, and principal component analysis is used to reduce the dimensionality of the multidimensional fault feature vectors to obtain dimensionality-reduced feature vectors. The dimensionality-reduced feature vectors are mapped to a predefined fault location coordinate system associated with the physical structure of the automatic tension control valve to generate fault location coordinates.

8. The method according to claim 1, characterized in that, Step S5 verifies the stability of the fault location coordinates through dynamic simulation, including: Based on the geometric structure and physical parameters of the automatic tension control valve, a three-dimensional virtual model including the main flow channel, auxiliary flow channels and connection relationships is constructed, and the fault location coordinate space is mapped to the internal position of the corresponding flow channel in the virtual model. Using real-time collected pressure data, flow data, and medium viscosity data as boundary conditions, a dynamic working condition simulation algorithm based on computational fluid dynamics principles is employed to simulate the dynamic fluid interaction process between the main flow channel and the auxiliary flow channel, and to obtain simulation results. Extract the flow field parameter sequence related to the fault location coordinates from the simulation results, calculate the statistical variance of the flow field parameter sequence, and define the variance value as the stability index; The stability index is compared with the preset stability conditions. If the stability index meets the preset stability conditions, the fault location coordinates are determined to be stable.

9. A fault diagnosis system for an automatic tension control valve of a scraper conveyor, used to implement the fault diagnosis method for an automatic tension control valve of a scraper conveyor as described in any one of claims 1-8, characterized in that, The system includes: The data acquisition module is used to acquire pressure and flow data of the main channel and auxiliary channel of the automatic tension control valve through a sensor array, and to extract the frequency domain features of the pressure and flow data using signal processing methods to obtain the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel. The anomaly analysis module is used to calculate the cross-correlation coefficient based on the pressure spectrum of the main channel and the flow spectrum of the auxiliary channel, determine whether the cross-correlation coefficient is lower than a preset threshold, and if so, determine that there is a parameter coupling anomaly between the main channel and the auxiliary channel, and generate an anomaly coupling index. The correction processing module is used to analyze the abnormal coupling index by combining the environmental interference data of the scraper conveyor, obtain the initial distribution range of the nonlinear flow characteristics, correct the initial distribution range by integrating the medium characteristic data, and determine whether the corrected distribution range deviates from the normal reference. If so, the correction deviation value is output. The preliminary positioning module is used to calculate the similarity score between the correction deviation value and the preset fault mode. If the similarity score is higher than the preset threshold, the corresponding fault type is determined. Based on the similarity score, fault positioning coordinates are generated through optimization and mapping processing. The stability verification module is used to map the fault location coordinates to the virtual model of the automatic tension control valve, and verify the stability of the fault location coordinates through dynamic simulation. If the stability index meets the preset conditions, a real-time warning signal containing the fault type and the fault location coordinates is generated.

Images