Mining belt conveyor fault early warning method, system and equipment and storage medium

Abstract

The invention discloses a fault early warning method, system and device for a mining belt conveyor and a storage medium, and the method comprises the steps that an associated sound wave signal and a motor current signal of the current belt conveyor are obtained, and the associated sound wave signal is acquired by a sensing optical fiber; signal conversion is conducted on the associated sound wave signals and the motor current signals, a sound spectrogram and a current spectrogram are obtained and input into a preset fault recognition model, a fault recognition result is obtained, a feature extraction module in the preset fault recognition model is constructed based on a parallel LSTM layer, and a feature fusion module obtains fusion features according to the current distribution weight; the fault classifier determines the fault category probability according to the fusion features, so that the fault recognition result is more accurate and reliable, the current allocation weight is dynamically generated based on the environmental parameters, and the adaptability of the model to different environments is enhanced; therefore, when the fault identification result meets the preset fault early warning condition, the early warning prompt is generated according to the fault identification result, and the reliability of fault early warning is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault early warning, in particular to a mine belt conveyor fault early warning method, system, device and storage medium. BACKGROUND

[0002] With the continuous development of the coal industry, coal production gradually moves towards automation and intelligence. The belt conveyor, as a key link in coal production, has important significance for ensuring production efficiency and worker safety.

[0003] However, the existing coal mine belt conveyor monitoring scheme has many shortcomings: traditional monitoring relies on manual inspection or local sensors. Manual inspection is low in efficiency and poor in real-time performance, and has coverage blind spots, making it difficult to discover potential faults in time. The monitoring range of local sensors is limited, and it is difficult to fully perceive the complex conditions of the entire belt conveyor line, which is prone to monitoring omissions. In addition, in complex underground environments, electrical sensors are easily affected by electromagnetic interference, further limiting their application effect, and it is difficult to meet the needs of the coal mine main transportation system for accurate and reliable early warning of belt conveyor faults. SUMMARY

[0004] The main purpose of the present application is to provide a mine belt conveyor fault early warning method, system, device and storage medium, which aims to solve the technical problems of low fault recognition accuracy of existing mine belt conveyors and difficulty in meeting real-time early warning needs.

[0005] To achieve the above purpose, the present application provides a mine belt conveyor fault early warning method, which comprises:

[0006] Obtaining the associated sound wave signal of the current belt conveyor and the motor current signal, the associated sound wave signal being collected by a sensing optical fiber installed on the roller frame of the current belt conveyor;

[0007] Respectively converting the associated sound wave signal and the motor current signal to obtain a sound spectrum graph and a current spectrum graph;

[0008] Inputting the sound spectrum graph and the current spectrum graph into a preset fault recognition model to obtain a fault recognition result, the preset fault recognition model comprising a feature extraction module, a feature fusion module and a fault classifier, the feature extraction module being constructed based on a parallel LSTM layer for feature extraction of model input, the feature fusion module being used for fusing feature extraction results according to a current allocation weight and obtaining fusion features, and the fault classifier being used for determining a fault class probability according to the fusion features, the current allocation weight being dynamically generated based on environmental parameters of an environment to which the current belt conveyor belongs;

[0009] When the fault identification result meets a preset fault early warning condition, an early warning prompt is generated according to the fault identification result.

[0010] In an embodiment, the step of inputting the sound spectrum diagram and the current spectrum diagram into a preset fault identification model to obtain a fault identification result comprises:

[0011] The feature extraction module is used to extract features from the sound spectrum diagram and the current spectrum diagram respectively to obtain sound features and current features;

[0012] The feature fusion module is used to obtain a current distribution weight and fuse the sound features and the current features according to the current distribution weight to obtain fused features;

[0013] The fault classifier is used to map the fused features to obtain a fault category probability and determine the fault category probability as the fault identification result.

[0014] In an embodiment, the preset fault identification model further comprises a dynamic weight generation module, and the dynamic weight generation module is constructed based on a multi-layer perception machine.

[0015] Before the step of obtaining a current distribution weight by the feature fusion module and fusing the sound features and the current features according to the current distribution weight to obtain fused features, the method further comprises:

[0016] An environment parameter of an environment to which the current belt conveyor belongs is obtained, and the environment parameter comprises a dust concentration and an electromagnetic intensity.

[0017] The dynamic weight generation module is used to perform nonlinear mapping according to the dust concentration and the electromagnetic intensity to obtain the current distribution weight.

[0018] The current distribution weight comprises a sound feature weight and a current feature weight.

[0019] Before the step of inputting the sound spectrum diagram and the current spectrum diagram into a preset fault identification model to obtain a fault identification result, the method further comprises:

[0020] A fault identification model to be trained is constructed based on a long short-term memory network structure and a multi-layer perception machine structure.

[0021] A belt conveyor fault data set is obtained, and the belt conveyor fault data set comprises a historical sound signal sequence, a historical circuit signal training sequence, and corresponding historical environment parameters and fault category labels.

[0022] inputting the historical sound wave signal sequence, the historical circuit signal sequence and the historical environment parameter into the fault identification model to be trained to obtain a predicted allocation weight and a predicted fault category;

[0023] inputting the predicted allocation weight, the predicted fault category and the fault category label into a preset loss function to obtain a current loss value;

[0024] updating parameters of the fault identification model to be trained according to the current loss value to obtain a preset fault identification model.

[0025] In an embodiment, the step of inputting the predicted allocation weight, the predicted fault category and the fault category label into a preset loss function to obtain a current loss value comprises:

[0026] inputting the predicted fault category and the fault category label into a preset classification loss function to obtain a current classification loss value;

[0027] inputting the predicted allocation weight into a preset weight prediction loss function to obtain a current weight loss value;

[0028] determining the current loss value according to the current classification loss value and the current weight loss value.

[0029] In an embodiment, the step of generating an early warning prompt according to the fault identification result when the fault identification result meets a preset fault early warning condition comprises:

[0030] determining a current fault category probability and a corresponding fault location according to the fault identification result;

[0031] obtaining temperature data of the current belt conveyor based on the sensing optical fiber;

[0032] determining whether the temperature data and the current fault category probability meet a preset fault early warning condition, the preset fault early warning condition comprising a probability condition and a temperature condition;

[0033] if yes, generating an early warning prompt according to the temperature data, the current fault category probability and the corresponding fault location.

[0034] In an embodiment, the sensing optical fiber on the roller carrier of the current belt conveyor comprises a first sensing optical fiber and a second sensing optical fiber;

[0035] The step of obtaining the associated sound wave signal of the current belt conveyor comprises:

[0036] respectively obtaining Rayleigh echo signals collected by the first sensing optical fiber and the second sensing optical fiber;

[0037] determine a first phase offset and a second phase offset according to the Rayleigh echo signals, and calculate a current phase deviation according to the first phase offset and the second phase offset;

[0038] When the current phase deviation is less than a preset deviation threshold, demodulate the Rayleigh echo signals collected by the first sensing optical fiber or the second sensing optical fiber to obtain the associated acoustic wave signals of the current belt conveyor.

[0039] In addition, to achieve the above object, the present application also proposes a mine belt conveyor fault early warning system, the system comprises:

[0040] A signal collection module is configured to collect the associated acoustic wave signals of the current belt conveyor and the motor current signals, wherein the associated acoustic wave signals are collected by the sensing optical fiber installed on the roller frame of the current belt conveyor.

[0041] A signal processing module is configured to perform signal conversion on the associated acoustic wave signals and the motor current signals respectively to obtain sound spectrum graphs and current spectrum graphs.

[0042] A fault identification module is configured to input the sound spectrum graphs and the current spectrum graphs into a preset fault identification model to obtain a fault identification result, wherein the preset fault identification model comprises a feature extraction module, a feature fusion module and a fault classifier, the feature extraction module is constructed based on a parallel LSTM layer and is configured to perform feature extraction on the model input, the feature fusion module is configured to fuse the feature extraction results according to a current distribution weight and obtain fused features, and the fault classifier is configured to determine a fault category probability according to the fused features, wherein the current distribution weight is dynamically generated based on the environmental parameters of the environment to which the current belt conveyor belongs.

[0043] A fault early warning module is configured to generate an early warning prompt according to the fault identification result when the fault identification result meets a preset fault early warning condition.

[0044] In addition, to achieve the above object, the present application also proposes a mine belt conveyor fault early warning device, the device comprises a memory, a processor and a mine belt conveyor fault early warning program stored on the memory and executable on the processor, and the mine belt conveyor fault early warning program is configured to implement the steps of the mine belt conveyor fault early warning method as described above.

[0045] In addition, to achieve the above object, the present application also proposes a storage medium, which is a computer readable storage medium, and the storage medium stores a mine belt conveyor fault early warning program, and the mine belt conveyor fault early warning program implements the steps of the mine belt conveyor fault early warning method as described above when executed by a processor.

[0046] The application discloses a mine belt conveyor fault early warning method, which comprises the following steps: acquiring an associated sound wave signal and a motor current signal of a current belt conveyor, wherein the associated sound wave signal is collected by a sensing optical fiber installed on a roller support of the current belt conveyor; performing signal conversion on the associated sound wave signal and the motor current signal respectively to obtain a sound spectrum graph and a current spectrum graph; inputting the sound spectrum graph and the current spectrum graph into a preset fault recognition model to obtain a fault recognition result, wherein the preset fault recognition model comprises a feature extraction module, a feature fusion module and a fault classifier, the feature extraction module is constructed based on a parallel LSTM layer and is used for feature extraction of model input, the feature fusion module is used for fusing feature extraction results according to a current distribution weight and obtaining fused features, and the fault classifier is used for determining a fault category probability according to the fused features, and the current distribution weight is dynamically generated based on environmental parameters of an environment to which the current belt conveyor belongs; and generating an early warning prompt according to the fault recognition result when the fault recognition result meets a preset fault early warning condition.

[0047] Since the application can comprehensively utilize two different types of data, namely the sound wave signal and the motor current signal, deeply excavate respective features through the feature extraction module, and fuse the features through the dynamic weight of the feature fusion module, the fault recognition result is more accurate and reliable, and the accuracy and timeliness of the fault early warning are effectively improved. Meanwhile, the current distribution weight is dynamically generated according to the environmental parameters, the influence of actual working conditions on fault features is fully considered, the adaptability of the model to different environments is enhanced, and the intelligent level and applicability of the fault early warning are further improved. BRIEF DESCRIPTION OF DRAWINGS

[0048] The drawings incorporated into the specification and forming a part thereof, illustrate embodiments consistent with the application and, together with the description, serve to explain the principles of the application.

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without any creative work.

[0050] Figure 1 A flowchart of a first embodiment of the mine belt conveyor fault early warning method of the application;

[0051] Figure 2 A flowchart of a second embodiment of the mine belt conveyor fault early warning method of the application;

[0052] Figure 3 A model structure diagram of the preset fault recognition model of the application;

[0053] Figure 4Fig. 1 is a flowchart of a third embodiment of the belt conveyor fault early warning method of the present application;

[0054] Figure 5 Fig. 2 is a schematic diagram of the module structure of the belt conveyor fault early warning system of the present application;

[0055] Figure 6 Fig. 3 is a schematic diagram of the structure of the belt conveyor fault early warning device of the present application.

[0056] The object, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0057] It should be understood that the specific embodiments described herein are merely intended to explain the technical solutions of the present application, and are not intended to limit the present application.

[0058] In order to better understand the technical solutions of the present application, the specific embodiments will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] The embodiment of the present application provides a belt conveyor fault early warning method, referring to Figure 1 , Figure 1 Fig. 1 is a flowchart of a first embodiment of the belt conveyor fault early warning method of the present application, in which the method comprises steps S10-S40:

[0060] Step S10: Obtain the associated sound wave signal and motor current signal of the current belt conveyor, wherein the associated sound wave signal is collected by a sensing optical fiber installed on the roller support of the current belt conveyor.

[0061] It should be noted that the execution subject of the present embodiment can be a computing electronic device with data processing, program running, and network communication, such as a fault early warning server, a fault monitoring computer, etc. The following examples use a fault early warning server (referred to as "server") as an example to describe the embodiments of the present application.

[0062] It should be understood that the current belt conveyor is the mine belt conveyor that needs to be monitored for faults, and the sensing optical fiber is an optical fiber sensor which can be installed on the roller support of the current belt conveyor. The probe light is injected into the sensing optical fiber from one end, and each position of the optical fiber will generate a back Rayleigh echo, which carries the sound wave field parameters of the optical fiber molecular group lattice. When the sound wave energy of the surrounding environment is transmitted to the glass lattice in the optical fiber, the Rayleigh back echo generated by the lattice will vibrate with the vibration frequency of the lattice.

[0063] The server can also be connected with a fiber optic acoustic channel demodulator, and the fiber optic acoustic channel demodulator is connected with the sensing optical fiber, so that the Rayleigh echo can be demodulated, the vibration energy difference is demodulated, the acoustic field characteristics of the vibration source are obtained, and the sound restoration on site is realized through a series of algorithms to obtain the associated acoustic signal of the current belt conveyor.

[0064] In addition, in order to ensure the reliability of the obtained associated acoustic signal, two sensing optical fibers, a first sensing optical fiber and a second sensing optical fiber, can also be arranged in advance on the carrier roller frame of the current belt conveyor. Then the Rayleigh echo signals collected by the first sensing optical fiber and the second sensing optical fiber can be obtained respectively, and the first phase offset and the second phase offset can be determined according to the Rayleigh echo signals.

[0065] Since the phase offset reflects the phase change of the signal in the transmission process due to the influence of external acoustic waves and other factors, the current phase deviation, that is, the difference between the signals collected by the two sensing optical fibers, can be calculated according to the first phase offset and the second phase offset, so as to determine the quality and reliability of the signal.

[0066] Specifically, if the current phase deviation is less than a preset deviation threshold, it means that the phase change of the signal is within an acceptable range. In this case, the Rayleigh echo signals collected by the first sensing optical fiber or the second sensing optical fiber can be demodulated, and the key information related to the acoustic wave can be extracted from the Rayleigh echo signals to obtain the associated acoustic signal of the current belt conveyor, which provides data support for subsequent fault warning and other analyses. Otherwise, the first phase offset and the second phase offset can be used to locate the specific position on the current belt conveyor to directly prompt the user.

[0067] It should be noted that the motor current signal can be collected by a flameproof current transformer arranged on the power cable of the current belt conveyor. The motor current signal can be a continuous time sequence, and is time-aligned with the aforementioned associated acoustic signal.

[0068] Step S20: respectively performing signal conversion on the associated acoustic signal and the motor current signal to obtain a sound spectrum graph and a current spectrum graph.

[0069] It should be understood that since the aforementioned collected associated acoustic signal and motor current signal are both time-domain signals, in order to facilitate subsequent model processing, they can be respectively converted into frequency domain to obtain a sound spectrum graph and a current spectrum graph.

[0070] In a specific implementation, the associated acoustic signal can be subjected to Fourier transform (FFT) to obtain the amplitude spectrum and the phase spectrum of the acoustic signal, and the amplitude spectrum can be displayed in a graphical manner to obtain the sound spectrum graph. Similarly, the current spectrum graph can be obtained.

[0071] Step S30: inputting the sound spectrum diagram and the current spectrum diagram into a preset fault recognition model to obtain a fault recognition result, the preset fault recognition model comprising a feature extraction module, a feature fusion module and a fault classifier, the feature extraction module being constructed based on a parallel LSTM layer and being used for feature extraction of model input, the feature fusion module being used for fusion of feature extraction results according to a current distribution weight and obtaining fusion features, and the fault classifier being used for determining a fault category probability according to the fusion features, the current distribution weight being dynamically generated based on an environment parameter of an environment to which the current belt conveyor belongs.

[0072] It should be noted that the preset fault recognition model can be divided into the feature extraction module, the feature fusion module and the fault classifier connected in sequence. The feature extraction module can be constructed based on a long short-term memory network (LSTM) and composed of two parallel LSTM layers, and is used for feature extraction of the input sound spectrum diagram and current spectrum diagram respectively, so as to obtain sound features and current features.

[0073] The feature fusion module can fuse the sound features and current features by feature weighting after receiving the sound features and current features, to obtain fusion features. The corresponding weights of the sound features and current features can be dynamically generated based on a preset weight generation mechanism. The weight generation mechanism can be dynamically adjusted based on the environment parameter, so as to fully consider the influence of actual working conditions on fault features (fusion features).

[0074] Therefore, in the foregoing process of collecting the associated sound wave signal and motor current signal, the environment parameter of the environment in which the current belt conveyor is located, such as dust concentration or electromagnetic interference degree in the environment, can also be collected, so that the current distribution weight is generated based on the weight generation mechanism.

[0075] Finally, the fault classifier can determine a fault category probability according to the fusion features after receiving the fusion features, that is, calculate the possibility of different types of faults of the belt conveyor. Therefore, by analyzing the fault category probability, it can be determined whether the belt conveyor has a fault, the position of the fault and the type of the fault.

[0076] Step S40: generating an early warning prompt according to the fault recognition result when the fault recognition result meets a preset fault early warning condition.

[0077] It should be noted that the preset fault early warning condition can be a preset fault probability threshold, which can be set for different fault types respectively. If the current fault category probability in the fault recognition result is greater than the fault probability threshold, it is determined that the current belt conveyor has an abnormality, and a fault abnormality prompt can be generated.

[0078] Further, considering that the sensing optical fiber can also realize fiber temperature measurement function, and the temperature anomaly of the belt conveyor is also related to the fault condition of the belt conveyor, step S40 specifically includes steps S401-S404:

[0079] Step S401: determining a current fault category probability and a corresponding fault position according to the fault identification result.

[0080] In a specific implementation, the fault identification result can be a probability value of each possible fault category. For example, the fault identification result can include that the probability of the output belt conveyor appearing “run-off” is 60%, the probability of the “idler roller jamming” is 30%, and the like. Meanwhile, based on the phase shift of the Rayleigh echo signal collected by the sensing optical fiber, the fault point position where the fault exists can be determined.

[0081] Step S402: obtaining temperature data of the current belt conveyor based on the sensing optical fiber.

[0082] It should be understood that the server can also be connected with a fiber temperature channel demodulator, which can perform intensity demodulation on the collected Rayleigh echo, capture the slow-varying component of the scattered light intensity of the Rayleigh echo, and then obtain the temperature data of the current belt conveyor based on the fiber temperature measurement technology.

[0083] Step S403: determining whether the temperature data and the current fault category probability satisfy a preset fault warning condition, the preset fault warning condition including a probability condition and a temperature condition.

[0084] It should be understood that the probability condition can be to set a threshold (such as 80%) of a fault category probability. When the probability of a fault category predicted by the model exceeds the threshold, it is considered that the possibility of the fault is large, and the probability condition is satisfied.

[0085] The temperature condition can be to set a temperature range or threshold. For example, when the temperature exceeds a certain upper limit or is lower than a certain lower limit, it is considered that the belt conveyor is in an abnormal temperature environment, which can exacerbate the occurrence of the fault or affect the performance of the equipment, so that the temperature condition is satisfied.

[0086] Step S404: if yes, generating a warning prompt according to the temperature data, the current fault category probability, and the corresponding fault position.

[0087] It should be understood that the temperature data and the fault category probability can be considered to satisfy the preset fault warning condition when they simultaneously satisfy the respective preset conditions. This comprehensive judgment mechanism can reduce the situation of false triggering of the warning due to a single factor, and improve the accuracy and reliability of the warning.

[0088] When the temperature data or the fault category probability meets the corresponding preset condition, the preset fault warning condition is considered to be met, real-time warning based on temperature or based on fault condition is realized, and flexibility of the warning is improved.

[0089] In a specific implementation, when the preset fault warning condition is met, the server can integrate the temperature data, the current fault category probability, and the corresponding fault location together to generate a detailed warning prompt, which is conveyed to relevant personnel in multiple ways, such as sound alarm, light blinking, short message notification, pop-up warning window on a monitoring system, and the like, to ensure that the warning information can be received and responded to by the staff in a timely manner.

[0090] The embodiment can comprehensively utilize the two different types of data, i.e., the sound wave signal and the motor current signal, to deeply mine respective features through the feature extraction module, and to make the fault recognition result more accurate and reliable through dynamic weight fusion of the feature fusion module, thereby effectively improving the accuracy and timeliness of the fault warning. Meanwhile, the current distribution weight is dynamically generated according to the environmental parameters, the influence of the actual working condition on the fault feature is fully considered, the adaptability of the model to different environments is enhanced, and the intelligent level and applicability of the fault warning are further improved.

[0091] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as the above-mentioned first embodiment can be referred to the above description, and will not be described in detail. On this basis, please refer to Figure 2 , Figure 2 The flowchart of the second embodiment of the mine belt conveyor fault warning method of the present application is shown.

[0092] In the embodiment, in order to specifically describe the application process of the preset fault model, step S30 further includes steps S301-S303.

[0093] Step S301: The feature extraction module is used to respectively extract features from the sound spectrum graph and the current spectrum graph to obtain sound features and current features.

[0094] It should be noted that the feature extraction module can include a first LSTM layer and a second LSTM layer, which are respectively used to receive the sound spectrum graph and the current spectrum graph for feature extraction.

[0095] In a specific implementation, the first LSTM layer can extract features from the input sound spectrum graph to obtain sound features, which can include, for example, frequency components, energy distribution, amplitude of specific frequency bands, and the like, and can reflect key information in the sound wave signal; the second LSTM layer can extract features from the input current spectrum graph to obtain current features, which can include frequency, harmonic components, energy distribution, and the like of the current signal, and can help to understand the motor operating state.

[0096] Step S302: Obtain a current allocation weight by the feature fusion module, and perform weighted fusion on the sound feature and the current feature according to the current allocation weight to obtain a fusion feature.

[0097] It should be noted that the current allocation weight can be dynamically generated based on a weight generation mechanism, and can also be generated by a dynamic weight generation module in the preset fault recognition model. The dynamic weight generation module can be constructed based on a multi-layer perception (MLP). Therefore, before step S302, steps S001-S002 are further included:

[0098] Step S001: Obtain environmental parameters of an environment to which the current belt machine belongs, the environmental parameters including dust concentration and electromagnetic intensity.

[0099] It should be understood that in the environment where the current belt machine is located, a dust concentration sensor and an EMI electromagnetic sensor can be pre-set to monitor the dust concentration and electromagnetic intensity in the environment in real time.

[0100] Step S002: Perform nonlinear mapping on the dust concentration and the electromagnetic intensity by the dynamic weight generation module to obtain the current allocation weight.

[0101] It should be understood that the dynamic weight generation module can perform nonlinear mapping on the dust concentration and the electromagnetic intensity by MLP. MLP is a neural network structure that can learn the complex nonlinear relationship between input and output. Since the current allocation weight includes a sound feature weight and a current feature weight, the dynamic weight generation module can be an MLP trained to give a higher value to the current feature weight when the dust concentration is high, and to give a higher value to the sound feature weight when the electromagnetic intensity is high.

[0102] In a specific implementation, the environmental parameters (dust concentration and electromagnetic intensity) monitored at the collection time of the associated sound wave signal and motor current signal are input to the MLP. The MLP can perform nonlinear mapping based on the environmental parameters to output the sound feature weight and the current feature weight, thereby adjusting the contribution ratio of the sound feature and the current feature in the feature fusion process.

[0103] Correspondingly, in the feature fusion module, the sound feature and the current feature can be weighted and fused according to the obtained sound feature weight and current feature weight to obtain a fusion feature. The fusion feature integrates information from both sound and current, and can more comprehensively reflect the running state of the belt machine.

[0104] Step S303: Perform feature mapping on the fusion feature by the fault classifier to obtain a fault category probability and determine a fault recognition result.

[0105] It should be understood that the fault category can include: belt misalignment, idler jamming, motor phase loss, and fire risk, etc. Therefore, the fault identification result can be represented by the output vector as: [probability of misalignment, probability of jamming, probability of phase loss, probability of fire].

[0106] In a specific implementation, the fault classifier can be composed of a fully connected layer and a sofmax function layer, so that the fused features can be mapped to different fault category probabilities and output as the fault identification result of the model.

[0107] In addition, you can refer to this place. Figure 3 The model structure of the fault identification model in this application is described below. Figure 3 This is a schematic diagram of the model structure of the fault identification model preset in this application.

[0108] Depend on Figure 3 It can be seen that the preset fault identification model can be divided into: feature extraction module, dynamic weight generation module, feature fusion module and fault classifier.

[0109] The feature module is connected to the feature extraction module, the dynamic weight generation module, and the fault classifier.

[0110] The acoustic spectrum and the current spectrum are respectively input into the first LSTM layer and the second LSTM layer of the feature extraction module to extract sound features and current features.

[0111] Environmental parameters (dust concentration, electromagnetic intensity) are input into the dynamic weight generation module, and the current assigned weights (weights w of acoustic features) are generated through nonlinear mapping by the MLP. s and the weight w of the current characteristics e );

[0112] The feature fusion module receives sound features and current features, and applies the previously generated assigned weights to the sound features F. audio and current characteristics F current Perform concat fusion to obtain the fused feature F. fused =F audio ·w s +F current ·w e .

[0113] After receiving the fused features, the fault classifier maps the fused features to fault category probabilities based on the fully connected layer and the softmax function layer, and outputs them as the fault identification result.

[0114] The preset fault recognition model in the embodiment can extract features from the sound spectrum graph and the current spectrum graph respectively and then perform weighted fusion, fully combining key information of both sound wave signals and current signals. Compared with single feature analysis, this can more comprehensively reflect the running state of the belt conveyor, thereby improving the accuracy of fault recognition. Furthermore, the weights are dynamically generated according to the dust concentration and electromagnetic intensity of the environment where the belt conveyor is located, so that the feature fusion process can better adapt to different working environments, which is conducive to widening the application range of the model.

[0115] Based on the first and second embodiments of the present application, the same or similar contents as the above embodiments one and two can be referred to the above introduction, and will not be described in detail hereinafter. On this basis, please refer to Figure 4 , Figure 4 for the flowchart of the third embodiment of the mine belt conveyor fault early warning method of the present application.

[0116] In the embodiment, in order to illustrate how to train the above preset fault recognition model, before step S30, steps S01-S05 are further included:

[0117] Step S01: A fault recognition model to be trained is built based on a long short-term memory network structure and a multi-layer perception machine structure.

[0118] It should be understood that the fault recognition module can include a feature extraction module, a dynamic weight generation module, a feature fusion module, and a fault classifier.

[0119] Therefore, regarding the feature extraction module, two parallel LSTM layers can be built for feature extraction of the sound spectrum graph and the current spectrum graph. Each LSTM layer can process sequence data and capture the dependence in time, thereby extracting sound features and current features.

[0120] Regarding the dynamic weight generation module, an MLP can be constructed for generating dynamic weights according to environmental parameters (dust concentration and electromagnetic intensity). The MLP includes an input layer, one or more hidden layers, and an output layer, which can perform nonlinear transformation on the input environmental parameters, and output sound feature weights and current feature weights.

[0121] Regarding the feature fusion module, a concat layer can be designed for weighted fusion of the sound features and current features from the feature extraction module according to the weights output by the dynamic weight generation module, to obtain fusion features.

[0122] Regarding the fault classifier, a fully connected layer and a softmax function layer can be used for mapping the fusion features to fault category probabilities, and outputting the probability of each fault category.

[0123] Step S02: obtaining a belt conveyor fault data set, the belt conveyor fault data set comprising: a historical acoustic signal sequence, a historical current signal training sequence, and corresponding historical environment parameters and fault category labels.

[0124] It should be understood that the belt conveyor fault data set can be composed of multiple samples, each sample including a historical acoustic signal sequence, a historical current signal training sequence, and corresponding historical environment parameters and fault category labels.

[0125] The historical acoustic signal sequence can be a sound spectrum sequence obtained by preprocessing and converting sound signals collected by a sensing optical fiber; the historical current signal sequence can be a current spectrum sequence obtained by preprocessing and converting current signals collected by a current sensor; the historical environment parameters can include environmental parameters such as dust concentration and electromagnetic intensity, corresponding to each signal sequence sample; and the fault category label can be a fault category label corresponding to each sample, indicating which fault type (belt deviation, roller jam, motor open phase, and fire risk) the sample belongs to.

[0126] In addition, before training the model based on the samples in the belt conveyor fault data set, the historical acoustic signal sequence, the historical current signal sequence, and the historical environment parameters can also be normalized to facilitate better learning by the model. The normalization method can use, for example, min-max normalization, Z-score normalization, etc., which is not limited in the present embodiment.

[0127] Step S03: inputting the historical acoustic signal sequence, the historical current signal sequence, and the historical environment parameters into the fault recognition model to be trained to obtain a predicted allocation weight and a predicted fault category.

[0128] It should be noted that when the historical acoustic signal sequence, the historical current signal sequence, and the historical environment parameters are input into the fault recognition model to be trained, the model can sequentially perform feature extraction, weight generation, feature fusion, and special fault classification steps, thereby outputting the predicted probability of each fault category, i.e., the predicted fault category, and simultaneously obtaining the predicted allocation weight generated by the dynamic weight generation module.

[0129] Step S04: inputting the predicted allocation weight, the predicted fault category, and the fault category label into a preset loss function to obtain a current loss value.

[0130] It should be noted that the preset loss function can be composed of a preset classification loss function and a preset weight prediction loss function, which can be represented as follows:

[0131] L total = L cls + λ·L weight

[0132] wherein L total is a preset loss function, L cls is a preset classification loss function, L weight is a preset weight prediction loss function, and λ is a weight coefficient preset by a user to balance the influence of the classification loss and the weight prediction loss.

[0133] It should be noted that the preset classification loss function can be used to measure the prediction accuracy of the model for the fault category, and therefore the classification loss (such as cross-entropy loss) can be calculated for the output (fault category probability distribution) of the fault prediction model and the real fault label, that is, the preset classification loss function can be represented as:

[0134]

[0135] wherein C is the total number of fault categories, y i is the encoding of the real fault category label, and p i is the probability of the predicted fault category output by the model.

[0136] It should also be noted that since the training target of the MLP can be to give the current feature a higher weight value when the dust concentration is high, and to give the sound feature a higher weight value when the electromagnetic intensity is high, the loss function can be defined based on the relationship between the expected weight and the environmental parameter.

[0137] Specifically, the expected relationship can be that when the dust concentration d d increases, the weight w e of the current feature also increases; when the electromagnetic intensity d e increases, the weight w s of the sound feature also increases.

[0138] Therefore, the following preset weight prediction loss function can be designed:

[0139]

[0140] wherein and are the current feature weight and the dust concentration of the previous sample; and are the sound feature weight and the electromagnetic intensity of the previous sample; α and β are weight coefficients for balancing the contributions of the two terms; and E[·] represents the expected value, which can be approximated by the average of the samples in a batch.

[0141] In a specific implementation, the predicted fault category and the fault category label can be input to the preset classification loss function to obtain the current classification loss value L cls; input the prediction assignment weight into a preset weight prediction loss function to obtain a current weight loss value L weight ; and determine a current loss value L total .

[0142] Step S05: performing parameter updating on the to-be-trained fault identification model according to the current loss value, to obtain a preset fault identification model.

[0143] It should be noted that the current loss value L total comprehensively reflects the overall performance of the model on the classification task and the weight prediction task, and can be used to guide parameter updating in the model training process to optimize the performance of the model.

[0144] In a specific implementation, the current loss value L total is used to perform back propagation on the model parameters, calculate the gradient, and update the model parameters using, for example, an Adam optimization algorithm to minimize the current loss value. Through the back propagation algorithm, the parameters of each layer in the model are adjusted to minimize the value of the loss function.

[0145] The above forward propagation, loss calculation, back propagation, and model parameter optimization processes are repeated for continuous iteration training until the model converges (i.e., the current loss value no longer significantly decreases) or the pre-set iteration round is met, the training is completed, and a trained preset fault identification model is obtained.

[0146] The present embodiment trains in combination with the classification loss and the weight prediction loss. The model not only can accurately identify the fault category, but also can dynamically adjust the weight according to the environmental parameters, so that the model has excellent generalization ability under complex working conditions, adapts to variable operating environments, and improves the accuracy and reliability of fault early warning in different operating environments.

[0147] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the mine belt conveyor fault early warning method of the present application. More forms of simple transformation based on this technical concept are within the protection scope of the present application.

[0148] In addition, the present application also provides a mine belt conveyor fault early warning system, as shown in Figure 5 , which is a module structure schematic diagram of the mine belt conveyor fault early warning system of the present application. Figure 5 As shown in

[0149] , the mine belt conveyor fault early warning system includes a signal acquisition module 501, a signal processing module 502, a fault identification module 503, and a fault early warning module 504. Figure 5

[0150] ​The signal acquisition module 501 is configured to acquire an associated sound wave signal of the current belt conveyor and a motor current signal, and the associated sound wave signal is acquired by a sensing optical fiber installed on a roller frame of the current belt conveyor.

[0151] The signal processing module 502 is configured to perform signal conversion on the associated sound wave signal and the motor current signal respectively to obtain a sound spectrum diagram and a current spectrum diagram.

[0152] The fault identification module 503 is configured to input the sound spectrum diagram and the current spectrum diagram into a preset fault identification model to obtain a fault identification result, the preset fault identification model comprises a feature extraction module, a feature fusion module and a fault classifier, the feature extraction module is constructed based on a parallel LSTM layer and is configured to perform feature extraction on model input, the feature fusion module is configured to fuse feature extraction results according to a current distribution weight and obtain fused features, and the fault classifier is configured to determine a fault category probability according to the fused features, and the current distribution weight is dynamically generated based on an environmental parameter of an environment to which the current belt conveyor belongs.

[0153] The fault warning module 504 is configured to generate a warning prompt according to the fault identification result when the fault identification result meets a preset fault warning condition.

[0154] The system of the embodiment can comprehensively utilize two different types of data, i.e., the sound wave signal and the motor current signal, deeply mine respective features through the preset fault identification model, and perform dynamic weight fusion, so that the fault identification result is more accurate and reliable, and the accuracy and timeliness of the fault warning are effectively improved. Meanwhile, the current distribution weight is dynamically generated according to the environmental parameter, the influence of actual working conditions on fault features is fully considered, the adaptability of the model to different environments is enhanced, and the intelligent level and applicability of the fault warning are further improved.

[0155] The application also provides a mine belt conveyor fault warning device, which comprises at least one processor and a memory in communication connection with the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the mine belt conveyor fault warning method in the above embodiment one.

[0156] The following refers to Figure 6 , Figure 6This is a schematic diagram of the structure of the mine conveyor belt fault early warning device of this application. The mine conveyor belt fault early warning device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), etc., as well as fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The illustrated fault warning device for mining belt conveyors is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0157] like Figure 6 As shown, the mine conveyor belt fault early warning device may include a processing unit 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the mine conveyor belt fault early warning device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the mine conveyor belt fault warning device to exchange data wirelessly or via wired communication with other devices. Although the figure shows a mine conveyor belt fault warning device with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented or possessed alternatively.

[0158] The mine belt conveyor fault early warning device provided in the application adopts the mine belt conveyor fault early warning method in the above embodiment, and can solve the technical problem of mine belt conveyor fault early warning. Compared with the prior art, the beneficial effects of the mine belt conveyor fault early warning device provided in the application are the same as those of the mine belt conveyor fault early warning method provided in the above embodiment, and other technical features in the mine belt conveyor fault early warning device are the same as those disclosed in the previous embodiment method, which will not be repeated here.

[0159] The application also provides a computer readable storage medium having computer readable program instructions (i.e. computer programs) stored thereon for executing the mine belt conveyor fault early warning method in the above embodiment.

[0160] The computer readable storage medium provided in the application may, for example, be a U disk, but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system or device, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection with one or more conductive wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present embodiment, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system or device. The program code contained on the computer readable storage medium can be transmitted by any suitable medium, including but not limited to electrical wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0161] The readable storage medium provided in the application is a computer readable storage medium, which stores computer readable program instructions (i.e. computer programs) for executing the above mine belt conveyor fault early warning method, and can solve the technical problem of the mine belt conveyor fault early warning method. Compared with the prior art, the beneficial effects of the computer readable storage medium provided in the application are the same as those of the mine belt conveyor fault early warning method provided in the above embodiment, which will not be repeated here.

[0162] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element preceded by "comprises a" does not, without more limitations, foreclose the existence of additional elements of the process, method, article, or system that includes the element.

[0163] The above-mentioned embodiment serial numbers are only for description, not representing the advantages and disadvantages of the embodiments, and they are only some embodiments of the present application, and do not limit the scope of the present application. Any equivalent structural transformation made by using the content of the present application specification and drawings, or direct / indirect application in other related technical fields is included in the protection scope of the present application.

Claims

1. A mine belt conveyor fault early warning method, characterized in that, The method comprises: obtaining the associated sound wave signal of the current belt conveyor and the motor current signal, wherein the associated sound wave signal is collected by a sensing optical fiber installed on the roller frame of the current belt conveyor; respectively converting the associated sound wave signal and the motor current signal to obtain a sound spectrum graph and a current spectrum graph; inputting the sound spectrum graph and the current spectrum graph into a preset fault recognition model to obtain a fault recognition result, wherein the preset fault recognition model comprises a feature extraction module, a feature fusion module and a fault classifier, the feature extraction module is constructed based on a parallel LSTM layer and is used for feature extraction of model input, the feature fusion module is used for fusing feature extraction results according to a current distribution weight and obtaining fused features, and the fault classifier is used for determining a fault category probability according to the fused features, and the current distribution weight is dynamically generated based on environmental parameters of an environment to which the current belt conveyor belongs; when the fault recognition result meets a preset fault warning condition, generating a warning prompt according to the fault recognition result.

2. The method of claim 1, wherein, The step of inputting the sound spectrum graph and the current spectrum graph into a preset fault recognition model to obtain a fault recognition result comprises: extracting features of the sound spectrum graph and the current spectrum graph through the feature extraction module to obtain sound features and current features; obtaining a current distribution weight through the feature fusion module and weighting and fusing the sound features and the current features according to the current distribution weight to obtain fused features; mapping the fused features through the fault classifier to obtain a fault category probability and determine the fault category probability as the fault recognition result.

3. The method of claim 2, wherein, The preset fault recognition model further comprises a dynamic weight generation module, and the dynamic weight generation module is constructed based on a multi-layer perceptron; Before the step of obtaining a current distribution weight through the feature fusion module and weighting and fusing the sound features and the current features according to the current distribution weight to obtain fused features, the step further comprises: obtaining environmental parameters of an environment to which the current belt conveyor belongs, wherein the environmental parameters comprise dust concentration and electromagnetic intensity; obtaining the current distribution weight through the dynamic weight generation module by nonlinear mapping according to the dust concentration and the electromagnetic intensity; The current distribution weight comprises a sound feature weight and a current feature weight.

4. The method of claim 3, wherein, Before the step of inputting the sound spectrum graph and the current spectrum graph into a preset fault recognition model to obtain a fault recognition result, the step further comprises: building a fault recognition model to be trained based on a long short-term memory network structure and a multi-layer perceptron structure; obtaining a belt conveyor fault data set, wherein the belt conveyor fault data set comprises a historical sound wave signal sequence, a historical circuit signal training sequence and corresponding historical environmental parameters and fault category labels; inputting the historical sound wave signal sequence, the historical circuit signal sequence and the historical environmental parameters into the fault recognition model to be trained to obtain a predicted distribution weight and a predicted fault category; inputting the predicted assignment weight, the predicted fault category and the fault category label into a preset loss function to obtain a current loss value; updating parameters of the to-be-trained fault identification model according to the current loss value to obtain a preset fault identification model.

5. The method of claim 4, wherein, The step of inputting the predicted assignment weight, the predicted fault category and the fault category label into a preset loss function to obtain a current loss value comprises: inputting the predicted fault category and the fault category label into a preset classification loss function to obtain a current classification loss value; inputting the predicted assignment weight into a preset weight prediction loss function to obtain a current weight loss value; determining a current loss value according to the current classification loss value and the current weight loss value.

6. The method of claim 1, wherein, The step of generating an early warning prompt according to the fault identification result when the fault identification result meets a preset fault early warning condition comprises: determining a current fault category probability and a corresponding fault location according to the fault identification result; obtaining temperature data of the current belt conveyor based on the sensing optical fiber; determining whether the temperature data and the current fault category probability meet a preset fault early warning condition, the preset fault early warning condition comprising a probability condition and a temperature condition; if yes, generating an early warning prompt according to the temperature data, the current fault category probability and the corresponding fault location.

7. The method of claim 1, wherein, The sensing optical fiber on the roller carrier of the current belt conveyor comprises a first sensing optical fiber and a second sensing optical fiber. The step of obtaining the associated acoustic signal of the current belt conveyor comprises: respectively obtaining Rayleigh echo signals collected by the first sensing optical fiber and the second sensing optical fiber; determining a first phase offset and a second phase offset according to the Rayleigh echo signals, and calculating a current phase deviation according to the first phase offset and the second phase offset; when the current phase deviation is less than a preset deviation threshold, demodulating the Rayleigh echo signals collected by the first sensing optical fiber or the second sensing optical fiber to obtain the associated acoustic signal of the current belt conveyor.

8. A mine belt conveyor failure early warning system characterized by, The system comprises: a signal acquisition module configured to obtain an associated acoustic signal of a current belt conveyor and a motor current signal, the associated acoustic signal being collected by a sensing optical fiber installed on a roller carrier of the current belt conveyor; a signal processing module configured to respectively perform signal conversion on the associated acoustic signal and the motor current signal to obtain a sound spectrum graph and a current spectrum graph; a fault identification module configured to input the sound spectrum graph and the current spectrum graph into a preset fault identification model to obtain a fault identification result, the preset fault identification model comprising a feature extraction module, a feature fusion module and a fault classifier, the feature extraction module being constructed based on a parallel LSTM layer and configured to perform feature extraction on model input, the feature fusion module being configured to fuse feature extraction results according to a current assignment weight to obtain fused features, and the fault classifier being configured to determine a fault category probability according to the fused features, the current assignment weight being dynamically generated based on environmental parameters of an environment to which the current belt conveyor belongs. The fault early warning module is configured to generate an early warning prompt according to the fault identification result when the fault identification result meets a preset fault early warning condition.

9. A mine belt conveyor failure early warning device characterized by, The device comprises a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the mine belt conveyor fault early warning method according to any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the mine belt conveyor fault early warning method according to any one of claims 1 to 7.

Images