Method for monitoring air tightness and internal pressure faults of explosion-proof housing of mining motor

Abstract

The invention provides a mining motor explosion-proof housing airtightness and internal pressure fault monitoring method, and relates to the technical field of mechanical equipment monitoring, and the method comprises the steps: arranging a piezoelectric sensing array at a sealing key point of an explosion-proof housing, obtaining a vibration signal generated in a motor operation process, and converting the vibration signal into a stress distribution matrix through a piezoelectric effect; and analyzing the stress distribution matrix by using singular value decomposition, and identifying the spatial distribution of the stress abnormal region. Acoustic tracer gas is injected into the shell, acoustic response signals of the piezoelectric sensor in different frequency bands are collected, an acoustic characteristic coefficient group is extracted through wavelet transform, and the mapping relation between the characteristic coefficient group and a leakage path is established. And generating a fault probability density function by using a recurrent neural network in combination with the stress anomaly region and the time sequence change rule of the feature coefficient group, and determining the space coordinates of the leakage point. According to the invention, the real-time evaluation of the airtight state of the explosion-proof housing and the accurate positioning of the leakage point can be realized, and the reliability and accuracy of the positioning result are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of mechanical equipment monitoring technology, and more specifically, to a method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor. Background Technology

[0002] As a crucial component ensuring the safe operation of mining motors in high-risk environments such as coal mines, the airtightness and internal pressure monitoring technology of explosion-proof enclosures has received widespread attention in recent years. Explosion-proof enclosures not only need to withstand complex external environmental influences (such as high temperature, high humidity, and vibration), but also need to ensure complete isolation between their interior and the outside world to prevent safety accidents caused by explosive gas leaks. Traditional methods for detecting the airtightness of explosion-proof enclosures mainly rely on static pressure testing, leak detectors, and periodic manual inspections. However, these methods lack real-time performance and automation, making it difficult to promptly detect leaks caused by factors such as vibration, thermal expansion and contraction, or seal aging during operation. Furthermore, existing detection technologies have low accuracy in identifying the specific location of leaks, often requiring complete disassembly of the equipment for manual inspection, severely impacting the continuous operating efficiency of mining motors. In recent years, with the development of acoustic detection, piezoelectric sensing, and artificial intelligence technologies, airtightness monitoring methods based on multi-sensor fusion have gradually attracted researchers' attention. These methods, by acquiring vibration, acoustic, or stress characteristic signals, can theoretically improve the accuracy and real-time performance of leak detection. However, under complex working conditions, there are still significant technical challenges in extracting high-quality characteristic signals and accurately locating leak points.

[0003] Existing technologies for monitoring the airtightness of explosion-proof enclosures for mining motors have the following main shortcomings: First, traditional pressure testing and leak detection methods are mostly offline static detection, which cannot monitor the explosion-proof enclosure in operation in real time, easily leading to delayed detection of potential leaks; second, signal acquisition methods based on single sensors are easily affected by noise interference under complex working conditions, making it difficult to accurately extract the characteristics of leak signals; third, existing technologies have weak spatial positioning capabilities for leak points and lack multi-dimensional analysis methods that combine stress distribution and acoustic signals, resulting in large positioning errors; fourth, existing data analysis methods mostly rely on traditional algorithms, such as simple threshold judgment or linear analysis, which cannot effectively handle the temporal evolution characteristics of stress and acoustic signals, making it difficult to accurately assess the airtightness of the explosion-proof enclosure. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention is proposed. This invention provides a method for monitoring the airtightness and internal pressure faults of the explosion-proof casing of mining motors, which can, to some extent, solve the problems of insufficient real-time performance, inability to accurately extract leakage signal features, and low spatial positioning accuracy of leakage points in existing technologies.

[0005] According to one aspect of the present invention, a method for monitoring the airtightness and internal pressure faults of an explosion-proof enclosure for a mining motor is provided, comprising:

[0006] A piezoelectric sensor array is deployed at the key sealing points of the explosion-proof housing of the mining motor to obtain the vibration characteristic matrix A of the sealing points during motor operation;

[0007] Based on the piezoelectric effect of the piezoelectric sensing array, the vibration feature matrix A is converted into a stress distribution matrix B, and the spatial distribution of stress anomaly regions is identified by the singular value decomposition of the stress distribution matrix B.

[0008] Acoustic tracer gas is injected into the explosion-proof enclosure, and the acoustic response curves of the piezoelectric sensing array in different frequency bands are collected. Wavelet transform is used to extract the characteristic coefficient group C in the acoustic response curve, and the mapping relationship between the characteristic coefficient group C and the leakage path is established.

[0009] Based on the spatial distribution of the stress anomaly region and the temporal evolution of the characteristic coefficient group C, a recurrent neural network is used to evaluate the airtightness in real time, output the fault probability density function D, and determine the spatial coordinates of the leak point based on the peak position of the probability density function D.

[0010] Furthermore, based on the piezoelectric effect of the piezoelectric sensing array, the amplitude information in the vibration characteristic matrix A of the sealing point is converted into stress value through the piezoelectric equations of the piezoelectric ceramic material;

[0011] Specifically, by combining the piezoelectric constitutive equation with the mechanical-electric coupling effect, and by solving the equations simultaneously and eliminating the electric field term, an explicit expression for the stress value is obtained.

[0012] Furthermore, considering the influence of temperature changes, a temperature compensation mechanism is adopted. A temperature compensation coefficient is constructed using first-order and second-order temperature coefficients, and the compensated stress value is calculated using the current temperature and reference temperature as parameters. The compensated stress data is then reconstructed into a two-dimensional stress distribution matrix B based on the spatial location of the sensor.

[0013] Furthermore, the stress distribution matrix B is decomposed into a left singular matrix, a singular value diagonal matrix, and a right singular matrix using the singular value decomposition method. By analyzing the distribution characteristics of the singular values, abnormal patterns in the stress field are identified.

[0014] Furthermore, the identification of abnormal patterns in the stress field is achieved by defining a sequence of singular value ratios, constructing a stress anomaly discrimination function, calculating the spatial gradient of the stress anomaly discrimination function, and marking an abnormal region when the spatial gradient exceeds the characteristic value. Combined with historical data, the monitoring values ​​are smoothed through a dynamic evaluation function to identify abnormal points and regions.

[0015] Furthermore, by injecting helium as an acoustic tracer gas and using the piezoelectric sensing array to collect response signals, the leak path can be accurately located.

[0016] Furthermore, wavelet transform is performed on the response signal to extract leakage features, including wavelet transform coefficients, scaling parameters, and translation parameters, to obtain a time-frequency feature matrix.

[0017] Furthermore, the wavelet transform uses the Morlet wavelet as the mother wavelet function to extract the time-frequency features of the signal and construct a time-frequency feature matrix;

[0018] Based on the time-frequency feature matrix, energy features, frequency band energy ratio, time-frequency entropy, and wavelet coefficient peak ratio are further extracted, and a standardized feature vector is calculated by combining the energy proportion of each frequency band and the normalized time-frequency energy distribution.

[0019] Furthermore, based on the standardized feature vectors, the similarity between feature vectors is calculated using a kernel function, and a support vector machine decision function is constructed using Euclidean distance and kernel width parameters;

[0020] Based on the decision function, combined with Lagrange multipliers, category labels, and bias terms, the kernel function results are weighted and combined to determine the classification of the leakage signal.

[0021] Furthermore, based on the extraction of stress anomaly regions and characteristic coefficients, an LSTM network is used to perform real-time assessment of the airtightness of the explosion-proof enclosure of the mining motor and locate the leakage point.

[0022] Compared with existing technologies, this invention deploys a piezoelectric sensor array at key sealing points of the explosion-proof enclosure of mining motors. Utilizing the piezoelectric effect, it acquires vibration characteristics and converts them into a stress distribution matrix, enabling real-time identification of the spatial distribution of stress anomaly areas and improving the sensitivity of leak detection. The introduction of acoustic tracer gas and the combination of wavelet transform to extract acoustic response features significantly enhance the accuracy and anti-interference capability of leak signal feature extraction. By analyzing the temporal evolution of stress distribution and acoustic features through a recurrent neural network, real-time assessment of airtightness and precise location of leak points are achieved. Furthermore, by monitoring the stability of the leak point probability density and the distribution patterns of surrounding features, the reliability and accuracy of the location results are effectively improved, thus providing a more intelligent and efficient monitoring method for the safe operation of explosion-proof enclosures of mining motors. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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. In the drawings:

[0024] Figure 1 This is a flowchart of a method for monitoring the airtightness and internal pressure faults of an explosion-proof enclosure for a mining motor according to an embodiment of the present invention. Detailed Implementation

[0025] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0026] As mentioned in the background section, existing technologies have two main problems: First, most methods for monitoring the airtightness of explosion-proof enclosures for mining motors rely on offline static detection, which cannot achieve real-time monitoring during operation and can easily lead to delayed detection of leaks. Second, existing detection technologies based on single sensors are susceptible to noise interference under complex working conditions and cannot accurately extract leak signal features, resulting in low detection accuracy and stability. Third, existing leak point location methods lack multi-dimensional data fusion techniques, resulting in poor spatial positioning accuracy of leak points and difficulty in quickly and accurately determining the leak location.

[0027] Example 1

[0028] like Figure 1 As shown, the method for monitoring the airtightness and internal pressure faults of the explosion-proof enclosure of a mining motor includes:

[0029] S1: A piezoelectric sensor array is installed at the key sealing point of the explosion-proof housing of the mining motor to obtain the vibration characteristic matrix A of the sealing point during the operation of the motor, wherein the vibration characteristic matrix A contains amplitude, frequency and phase information.

[0030] At key sealing points of the explosion-proof enclosure for mining motors, based on the structural characteristics of the enclosure, multiple key sealing points are selected, including flange connections, cable entry devices, observation windows, and ventilated components. At each key sealing point, four piezoelectric sensors are evenly distributed circumferentially in a matrix arrangement, with a spacing of 2cm-5cm between adjacent sensors. Each piezoelectric sensor is fixed to the sealing surface using an elastic adhesive with a Shore hardness of 60-80 degrees to ensure tight adhesion between the sensor and the sealing surface. Each piezoelectric sensor is made of piezoelectric ceramic material with a piezoelectric constant of not less than 400 pC / N and an electromechanical coupling coefficient greater than 0.65.

[0031] During motor operation, the piezoelectric sensor array continuously collects vibration signals from the sealing point. The sampling frequency is set to 1kHz, and the sampling duration is 60s. The vibration signal is conditioned by a preamplifier with a signal amplification factor of 100x and a bandpass filter range of 10Hz-5kHz, forming a vibration signal data stream. Wavelet packet decomposition and Hilbert transform are performed on the vibration signal data stream to extract the amplitude envelope, instantaneous frequency, and phase difference between adjacent sensor signals at each sampling point. The amplitude envelope reflects the stress change trend at the sealing point, the instantaneous frequency characterizes the dynamic characteristics of the sealing point, and the phase difference is used to determine the vibration propagation direction.

[0032] The features are organized into an m×n dimensional vibration feature matrix A according to the time series, where m represents the number of sensors and n represents the dimension of the features. Each row of the vibration feature matrix A corresponds to the complete feature set of a sensor, and each column corresponds to the feature values ​​of different sensors at the same time. The vibration feature matrix A is uploaded to the data processing unit in real time through a wireless transmission module, and the transmission uses the AES-256 encryption algorithm to ensure data security.

[0033] S2: Based on the piezoelectric effect of the piezoelectric sensing array, the vibration feature matrix A of the sealing point is converted into a stress distribution matrix B, and the spatial distribution of the stress anomaly region is identified by the singular value decomposition of the stress distribution matrix B.

[0034] Based on the piezoelectric effect of the piezoelectric sensing array, the amplitude information in the vibration characteristic matrix A of the sealing point is converted into stress values ​​using the piezoelectric equations of the piezoelectric ceramic material. Specifically, for the first... Each sensor at time Output charge Based on the piezoelectric constitutive equation, we can obtain:

[0035]

[0036] in, The piezoelectric stress constant is This is the stress value. Where is the dielectric constant. denoted as electric field strength.

[0037] Meanwhile, considering the mechanical-electric coupling effect of piezoelectric ceramics, the stress and strain satisfy the following:

[0038]

[0039] in, It is the elasticity compliance coefficient. This is the dependent variable.

[0040] By solving the above equations simultaneously and eliminating the electric field term, we obtain an explicit expression for the stress value:

[0041]

[0042] Furthermore, considering that the properties of piezoelectric materials change with temperature, a temperature compensation mechanism is introduced, wherein the temperature compensation coefficient is... Represented as:

[0043]

[0044] in, The current temperature. For reference temperature, and These are the first-order and second-order temperature coefficients, respectively.

[0045] The stress value after temperature compensation is:

[0046]

[0047] The compensated stress data is reconstructed into a two-dimensional stress distribution matrix B according to the spatial location of the sensor. Due to the spatial attenuation effect during stress transmission, a spatial weighting function is introduced. :

[0048]

[0049] in, The coordinates of the sensor relative to the center of the seal are: The characteristic radius, This is the attenuation coefficient.

[0050] The stress distribution matrix elements after incorporating spatial weights are represented as follows:

[0051]

[0052] in, The first element in the reconstructed stress distribution matrix represents the... OK The element values ​​of the column, This represents the stress value after temperature compensation.

[0053] To extract anomalous features from the stress distribution matrix, singular value decomposition is used to analyze the stress distribution matrix. Perform singular value decomposition:

[0054]

[0055] in, for × 3D stress distribution matrix for × Left-singular matrix It is a singular value diagonal matrix. for × A right-singular matrix.

[0056] Furthermore, the distribution characteristics of singular values ​​are analyzed to identify anomalous patterns in the stress field.

[0057] Define the singular value ratio sequence as:

[0058]

[0059] in, For the first A singular value, For the first +1 singular value.

[0060] Construct a stress anomaly discrimination function based on the singular value ratio sequence. :

[0061]

[0062] in, The number of valid singular values, and These are the left and right singular vectors, respectively.

[0063] Through calculation The spatial gradient is used to identify anomalous regions in the stress field. The spatial gradient is represented as:

[0064]

[0065] Meanwhile, to improve the stability of the recognition results, a dynamic evaluation function with a memory effect is introduced:

[0066]

[0067] in, The forgetting factor has a value range of [0,1]. This represents the evaluation value at the current moment. This represents the evaluation value at the previous moment.

[0068] The final spatial distribution characteristics of the stress anomaly region are obtained through The spatiotemporal evolution characteristics were determined, specifically by calculating the spatiotemporal evolution characteristics of each spatial point using 30 consecutive days of historical data. Mean μ and standard deviation σ. When monitoring points When a value exceeds μ+3σ for five consecutive sampling periods, the point is marked as an initial outlier. For each initial outlier, its... The growth rate is measured, and a point is identified as an outlier when the growth rate exceeds 0.5 / min and the duration exceeds 10 minutes. Spatially, when the number of adjacent outliers exceeds 3 and the distance between them is... When the value difference is less than 20%, these point groups are defined as an abnormal region.

[0069] For confirmed abnormal areas, track them. The trend of value change: If any point in the region If the value increases by more than 50% within 15 minutes, it is considered an abnormal surge; if the value at the edge of the region... When the growth rate of the value relative to the center point exceeds 30%, it is judged as abnormal diffusion.

[0070] By calculating the rate of change of the centroid coordinates and area of ​​the anomalous region over time, the direction and speed of the anomalous region's development can be determined. When the value decreases continuously for 30 minutes and falls below μ+2σ, the region is marked as an anomalous regression. The final output shows the specific coordinate range, area size, development direction, and evolution rate of the anomalous region.

[0071] S3: Inject acoustic tracer gas into the explosion-proof enclosure, collect the acoustic response curves of the piezoelectric sensing array in different frequency bands, extract the characteristic coefficient group C in the acoustic response curve using wavelet transform, and establish the mapping relationship between the characteristic coefficient group C and the leakage path.

[0072] To further determine the specific location and characteristics of the leakage path, a specific tracer gas was injected into the explosion-proof enclosure, and the distributed detection capability of the piezoelectric sensor array was combined to achieve precise location and characterization of the leakage channel.

[0073] Specifically, helium is selected as the acoustic tracer gas. During the injection process, the inflation pressure is strictly controlled to be 1.2 times the standard atmospheric pressure to ensure sufficient pressure difference to drive gas leakage and avoid excessive pressure from causing additional damage to the sealing structure.

[0074] The response signal acquired by the piezoelectric sensing array is represented as:

[0075]

[0076] in, For the first The time-domain signal collected by the sensor Characterizing the instantaneous amplitude of a signal, Includes phase information, This represents the environmental noise superimposed on the valid signal.

[0077] Based on the frequency characteristics of the response signal, the sampling frequency is set in the range of 20kHz to 100kHz and evenly divided into 16 frequency bands.

[0078] The acquired acoustic signal is processed by wavelet transform to extract the leakage features contained in the signal, which are represented as follows:

[0079]

[0080] in, These are the wavelet transform coefficients. The scaling parameter controls the wavelet's scaling, corresponding to the frequency resolution. These are translation parameters that control the time position of the wavelet. This is the conjugate form of the mother wavelet function. It is the energy normalization factor.

[0081] Furthermore, ( Morlet wavelet is selected as the mother wavelet function:

[0082]

[0083] in, The normalization coefficient is... Gaussian envelope, controlling for temporal localization properties. For the center frequency, It is a time variable.

[0084] The time-frequency characteristic matrix obtained by wavelet transform is expressed as follows:

[0085]

[0086] Feature coefficients are extracted based on the time-frequency feature matrix, including: energy feature C1, frequency band energy ratio C2, time-frequency entropy C3, and wavelet coefficient peak ratio C4.

[0087] Specifically,

[0088]

[0089]

[0090]

[0091]

[0092] in, Indicates the first Energy percentage of each frequency band Indicates the first Energy of each frequency band Indicates the frequency band number. This represents the normalized time-frequency energy distribution. This represents the maximum absolute value of the wavelet coefficients. This represents the root mean square value of the wavelet coefficients.

[0093] Furthermore,

[0094]

[0095] By combining all the extracted feature coefficients into a 19-dimensional feature vector This achieves a complete description of the leakage signal characteristics, and each leakage sample will form a standardized feature vector, represented as:

[0096]

[0097] Based on the constructed feature vectors, the similarity between feature vectors is calculated using the kernel function K, expressed as:

[0098]

[0099] in, , Representing two eigenvectors, The parameter representing the width of the kernel function. This represents the Euclidean distance between eigenvectors.

[0100] Based on the kernel function, a decision function for the support vector machine is constructed. The results of the kernel function are then weighted and combined using the decision function, as follows:

[0101]

[0102] in, It is a classification decision value. These are the Lagrange multipliers obtained through optimization. It is the class label (±1) of the training sample. It is a bias term. It represents the number of support vectors.

[0103] By calculating the kernel function values ​​of the sample to be classified and all support vectors, the classification result is finally determined based on the sign of the decision function. If the output of the decision function is greater than zero, it indicates that the currently detected leak is located at the connection of the inlet device of the explosion-proof enclosure. This type of leak is usually caused by aging of the sealing ring, excessive installation gap, or loose fasteners, and has a regular annular leakage channel. Its acoustic characteristics show energy concentration in the high-frequency band. If the output of the decision function is less than zero, it indicates that the leak is located at the flange connection of the explosion-proof enclosure. This type of leak is mostly caused by deformation of the sealing surface, uneven surface roughness, or improper assembly stress, forming an irregular planar leakage channel. Its acoustic characteristics show significant energy distribution in the mid-to-low frequency band.

[0104] S4: Based on the spatial distribution of the stress anomaly region and the temporal evolution of the characteristic coefficient group C, a recurrent neural network is used to evaluate the airtightness in real time, output the fault probability density function D, and determine the spatial coordinates of the leakage point based on the peak position of the probability density function D.

[0105] Based on stress anomaly region identification and feature coefficient extraction, in order to achieve accurate location of leakage points and real-time status assessment, a recurrent neural network is further adopted to perform real-time assessment and accurate location of the airtightness status of the explosion-proof enclosure of the mining motor.

[0106] First, an input feature matrix is ​​constructed, consisting of two parts: one is the spatial distribution features of the stress anomaly region, which uses a p×q dimensional matrix to represent the stress distribution state at each detection time, where the stress value in the anomaly region is significantly higher than that in the background region; the other is the temporal data of the feature coefficient group C, which records the feature evolution process within n consecutive time windows, forming an n×19 dimensional temporal feature matrix (19 being the feature dimension).

[0107] The recurrent neural network uses a long short-term memory (LSTM) network structure, which contains three layers of LSTM units, each containing 128 neurons.

[0108] The output layer of the recurrent neural network is a Gaussian mixture model, used to generate the fault probability density function D:

[0109]

[0110] in, The number of Gaussian components. These are time-varying weighting coefficients. and They represent the first The mean vector and covariance matrix of each Gaussian component. This represents the Gaussian distribution function.

[0111] When the explosion-proof enclosure is in normal condition, the failure probability density function exhibits a smooth distribution; when leakage occurs, the probability density function forms a significant peak at the leakage point.

[0112] Specifically, if in position ( , The probability density value at point () satisfies:

[0113]

[0114]

[0115] in, This is the threshold coefficient, initially set to 3. Let be the average probability density at the current moment, then ( , The spatial coordinates of the leak point were determined.

[0116] To improve the reliability of the location, once a point is initially identified as a leak point, the probability density changes of that point are continuously monitored over eight time windows. If the spatial coordinates of the point drift within this time period do not exceed 2 mm, it is preliminarily confirmed as a stable leak point. Simultaneously, the probability density distribution pattern within a 10 mm × 10 mm area surrounding the point is analyzed. If this distribution exhibits a gradual decay from the center outwards, and the decay curve conforms to the expected physical model, it further supports the leak point identification result.

[0117] The recurrent neural network was trained using a large amount of historical data, including normal operation data under different working conditions and various leakage and fault data. The Adam optimizer was used during training, with an initial learning rate of 0.001, dynamically adjusted using cosine annealing. To prevent overfitting, Dropout layers (dropout rate 0.3) and L2 regularization (coefficient 0.0001) were added to the network. The training batch size was set to 64, with a total of 200 epochs, and an early stopping strategy was used to avoid overtraining.

[0118] After completing basic training, the model was fine-tuned using calibration data collected in the laboratory. During the adjustment process, the model parameters were continuously optimized to better adapt to the complex environments of actual working conditions, providing crucial protection for the safe operation of industrial equipment.

[0119] Example 2

[0120] The explosion-proof enclosure to be tested is fixed on a dedicated testing platform, and environmental parameters such as the current ambient temperature (28.5°C), humidity (65%), and atmospheric pressure (0.1 MPa) are recorded. After confirming that all piezoelectric sensors are working properly, the signal acquisition system is calibrated.

[0121] Subsequently, piezoelectric sensor arrays were installed at the key sealing points of the explosion-proof enclosure, including three sensors with uniform spacing on each of the upper and lower flange surfaces, one sensor on each of the left and right shaft seals, and one sensor at each of the three cable entry points. The sampling frequency of all sensors was set to 100Hz, and the spatial resolution was 10mm.

[0122] After the sensor deployment is completed, the data acquisition system is activated to obtain the initial state signals of each sensor and to collect the vibration characteristic matrix A of the sealing points during motor operation. Based on the piezoelectric effect of the piezoelectric sensor, the collected vibration characteristics are converted into a stress distribution matrix B. Singular value decomposition is performed on the stress distribution matrix to identify stress anomaly regions. Then, a specific concentration of acoustic tracer gas is injected into the explosion-proof enclosure, with the gas pressure controlled at 0.15 MPa. Simultaneously, the acoustic response curves of each sensor at different frequency bands are acquired, and wavelet transform is used to extract the characteristic coefficient set C.

[0123] A recurrent neural network is used to process the collected sensor data in real time, establish a mapping relationship between the feature coefficient set C and the leakage path, generate a fault probability density function D, and finally determine the precise spatial coordinates of the leakage point by analyzing the peak position of the probability density function.

[0124] Analysis revealed a leak at the upper flange face coordinates (450mm, 0mm), with a leakage probability of 92%, a leakage rate of 2.3×10⁻³ Pa·m³ / s, and a local pressure of 0.15MPa. Simultaneously, a leak was also detected at the right shaft seal coordinates (620mm, 150mm), with a leakage probability of 78%, a leakage rate of 1.8×10⁻³ Pa·m³ / s, and a local pressure of 0.12MPa.

[0125] The pressure drop rate of the entire system was 0.15 MPa / h, and the location reliability reached 95.2%. Based on the test results, it was determined that the leakage problem on the upper flange surface needed to be addressed promptly, and it was recommended to closely monitor the changing trend of the sealing condition at the right shaft seal.

[0126] In summary, the method for monitoring the airtightness and internal pressure faults of the explosion-proof enclosure of mining motors based on embodiments of the present invention has been clarified. By deploying a piezoelectric sensor array at key sealing points of the explosion-proof enclosure, the vibration characteristics are acquired using the piezoelectric effect and converted into a stress distribution matrix, enabling real-time identification of the spatial distribution of stress anomaly areas and improving the sensitivity of leak detection. The introduction of acoustic tracer gas and the extraction of acoustic response features using wavelet transform significantly improve the accuracy and anti-interference capability of leak signal feature extraction. The analysis of the temporal evolution of stress distribution and acoustic features using a recurrent neural network enables real-time assessment of airtightness and precise location of leak points. Furthermore, by monitoring the stability of the probability density of leak points and the distribution patterns of surrounding features, the reliability and accuracy of the location results are effectively improved, thus providing a more intelligent and efficient monitoring method for the safe operation of explosion-proof enclosures of mining motors.

Claims

1. A method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor, characterized in that, include: A piezoelectric sensor array is deployed at the key sealing points of the explosion-proof housing of the mining motor to obtain the vibration characteristic matrix A of the sealing points during motor operation; Based on the piezoelectric effect of the piezoelectric sensing array, the vibration feature matrix A is converted into a stress distribution matrix B, and the spatial distribution of stress anomaly regions is identified by the singular value decomposition of the stress distribution matrix B. Acoustic tracer gas is injected into the explosion-proof enclosure, and the acoustic response curves of the piezoelectric sensing array in different frequency bands are collected. Wavelet transform is used to extract the characteristic coefficient group C in the acoustic response curve, and the mapping relationship between the characteristic coefficient group C and the leakage path is established. Based on the spatial distribution of the stress anomaly region and the temporal evolution of the characteristic coefficient group C, a recurrent neural network is used to evaluate the airtightness in real time, output the fault probability density function D, and determine the spatial coordinates of the leak point based on the peak position of the probability density function D.

2. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 1, characterized in that, Based on the piezoelectric effect of the piezoelectric sensing array, the amplitude information in the vibration characteristic matrix A of the sealing point is converted into stress value through the piezoelectric equations of the piezoelectric ceramic material. Specifically, by combining the piezoelectric constitutive equation with the mechanical-electric coupling effect, and by solving the equations simultaneously and eliminating the electric field term, an explicit expression for the stress value is obtained.

3. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 2, characterized in that, Taking into account the influence of temperature changes, a temperature compensation mechanism is adopted. The temperature compensation coefficient is constructed by using first-order and second-order temperature coefficients, and the compensated stress value is calculated using the current temperature and reference temperature as parameters. The compensated stress data is reconstructed into a two-dimensional stress distribution matrix B based on the spatial location of the sensor.

4. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 3, characterized in that, The stress distribution matrix B is decomposed into a left singular matrix, a singular value diagonal matrix, and a right singular matrix using the singular value decomposition method. Anomalies in the stress field are identified by analyzing the distribution characteristics of the singular values.

5. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 4, characterized in that, The identification of abnormal patterns in the stress field is achieved by defining a sequence of singular value ratios, constructing a stress anomaly discrimination function, calculating the spatial gradient of the stress anomaly discrimination function, and marking an abnormal region when the spatial gradient exceeds the characteristic value. Combined with historical data, the monitoring values ​​are smoothed through a dynamic evaluation function to identify anomaly points and regions.

6. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 1, characterized in that, By injecting helium as an acoustic tracer gas and using the piezoelectric sensing array to collect response signals, the leak path can be accurately located.

7. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 6, characterized in that, The response signal is subjected to wavelet transform to extract leakage features, including wavelet transform coefficients, scaling parameters and translation parameters, to obtain a time-frequency feature matrix.

8. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 7, characterized in that, The wavelet transform uses Morlet wavelet as the mother wavelet function to extract the time-frequency features of the signal and construct a time-frequency feature matrix. Based on the time-frequency feature matrix, energy features, frequency band energy ratio, time-frequency entropy, and wavelet coefficient peak ratio are further extracted, and a standardized feature vector is calculated by combining the energy proportion of each frequency band and the normalized time-frequency energy distribution.

9. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 8, characterized in that, Based on the standardized feature vectors, the similarity between feature vectors is calculated using a kernel function, and a support vector machine decision function is constructed using Euclidean distance and kernel width parameters. Based on the decision function, combined with Lagrange multipliers, category labels, and bias terms, the kernel function results are weighted and combined to determine the classification of the leakage signal.

10. The method for monitoring the airtightness and internal pressure faults of the explosion-proof housing of a mining motor according to claim 1, characterized in that, Based on the extraction of stress anomaly regions and characteristic coefficients, an LSTM network is used to perform real-time assessment of the airtightness of the explosion-proof enclosure of mining motors and locate leakage points.

Images