Method and device for determining a belt conveyor failure and belt conveyor system

Abstract

The application provides a belt conveyor fault determination method and device and a belt conveyor system. The method comprises: acquiring ultrasonic data, which is data of ultrasonic waves reflected by the belt conveyor after the ultrasonic waves are emitted to the belt conveyor; constructing a target model, which is obtained by training a plurality of sets of training data by using an ELM algorithm and a VPMCD algorithm, each set of training data comprising historical ultrasonic data acquired in a historical time period and a first historical determination result corresponding to the historical ultrasonic data, the first historical determination result being used to represent a result of whether the historical belt conveyor is faulty; and inputting the ultrasonic data into the target model to obtain a first determination result corresponding to the ultrasonic data, wherein the first determination result is used to represent a result of whether the belt conveyor is faulty according to the target model. The scheme solves the problem of low accuracy of fault detection of the belt conveyor in the prior art.

Description

Technical Field

[0001] This application relates to the field of belt conveyor fault detection technology, and more specifically, to a method, apparatus, computer-readable storage medium, and belt conveyor system for determining belt conveyor faults. Background Technology

[0002] Currently, it is often necessary to dynamically identify belt conveyors in mines to determine their faults. Neural networks are often used to extract features for defect identification. However, neural networks cannot converge quickly and are prone to local minima. Furthermore, a large number of samples are required during training, which results in low accuracy in belt conveyor fault detection. Summary of the Invention

[0003] The main objective of this application is to provide a method, apparatus, computer-readable storage medium, and belt conveyor system for determining belt conveyor faults, so as to at least solve the problem of low accuracy in fault detection of belt conveyors in the prior art.

[0004] To achieve the above objectives, according to one aspect of this application, a method for determining belt conveyor faults is provided, comprising: acquiring ultrasonic data, wherein the ultrasonic data is data of ultrasonic waves reflected by the belt conveyor after ultrasonic waves are emitted to it; constructing a target model, wherein the target model is obtained by training multiple sets of training data using an ELM algorithm and a VPMCD algorithm, each set of training data including historical ultrasonic data acquired within a historical time period and a first historical determination result corresponding to the historical ultrasonic data, wherein the first historical determination result is used to characterize whether the historical belt conveyor was faulty; inputting the ultrasonic data into the target model to obtain the first determination result corresponding to the ultrasonic data, wherein the first determination result is used to characterize the result of determining whether the belt conveyor is faulty based on the target model.

[0005] Optionally, before constructing the target model, the method further includes: constructing an initial model, wherein the initial model is obtained by training multiple sets of training data using the VPMCD algorithm, each set of training data including the historical ultrasonic data acquired within the historical time period and a second historical determination result corresponding to the historical ultrasonic data, wherein the second historical determination result is used to characterize whether the historical belt conveyor was faulty; and optimizing the initial model using the ELM algorithm to obtain the target model.

[0006] Optionally, the initial model is optimized using the ELM algorithm, including: obtaining a training set, wherein the training set is T = (x1, y1), (x2, y2), ..., (x n ,y nThe input samples are mapped to the target feature space through the objective function to obtain the first matrix, where the input samples are X = x1, x2, ..., x... n The first matrix is ​​H = [h ij ] K×n Where K×n represents the dimension of the matrix, K represents the number of rows in the matrix, n represents the number of columns in the matrix, and h ij =g(z) k ), g(z) k ) represents the activation function of the k-th neuron in the hidden layer, and the objective function is either the sigmoid function or the ReLU function; the connection weights between the hidden layer and the output layer are randomly initialized and converted into matrix form to obtain the output data of the output layer, wherein the connection weights are W. K×m m is the number of neurons in the output layer, and the output data of the output layer is Y = [y i ] m×n , Y = W·H.

[0007] Optionally, the initial model is optimized using the ELM algorithm, including: calculating the fitting error between the output layer and the training data using the least squares method; and obtaining the residual sum of squares, wherein the residual sum of squares is...

[0008]

[0009] y i Let t represent an n-dimensional vector. i Represent an n-dimensional vector; convert the sum of squared residuals into a second matrix, where the second matrix is...

[0010]

[0011] Where, T = [t ij ] m×n T represents the target vector of the output layer, and the weights of the output layer are W = (H T H) - 1 H T H.

[0012] Optionally, the initial model is optimized using the ELM algorithm, including: obtaining the bias function between the hidden layer and the output layer; randomly initializing the bias function and optimizing the bias function using a target optimization algorithm to obtain the optimized bias function, wherein the target optimization algorithm is a learning rate decay algorithm or an adaptive learning rate algorithm.

[0013] Optionally, after optimizing the initial model using the ELM algorithm, the method further includes: obtaining the sample x to be determined by the following formula:

[0014]

[0015] The sample to be determined is mapped to a hidden layer through an activation function; the product of the output of the mapped hidden layer and the connection weight from the hidden layer to the output layer is calculated to obtain the calculation result, wherein the calculation result is y = W·h; the calculation result is classified to determine the category of the sample to be determined.

[0016] Optionally, after inputting the ultrasonic data into the target model to obtain the determination result corresponding to the ultrasonic data, the method further includes: obtaining a reference determination result, wherein the reference determination result is the result of fault identification of the belt conveyor standard; inputting the ultrasonic data into the initial model to obtain a second determination result corresponding to the ultrasonic data, wherein the second determination result is used to characterize the result of determining whether the belt conveyor is faulty based on the initial model; calculating a first error value between the first determination result and the reference determination result, and calculating a second error value between the second determination result and the reference determination result; displaying the first error value and the second error value in a target display format on the display interface, wherein the target display format is a chart display and / or a table display.

[0017] According to another aspect of this application, a device for determining belt conveyor faults is provided, comprising: a first acquisition unit for acquiring ultrasonic data, wherein the ultrasonic data is data of ultrasonic waves reflected by the belt conveyor after ultrasonic waves are emitted to it; a first construction unit for constructing a target model, wherein the target model is trained using multiple sets of training data through an ELM algorithm and a VPMCD algorithm, each set of training data including historical ultrasonic data acquired within a historical time period and a first historical determination result corresponding to the historical ultrasonic data, wherein the first historical determination result is used to characterize whether the belt conveyor was faulty in the past; and a determination unit for inputting the ultrasonic data into the target model to obtain the first determination result corresponding to the ultrasonic data, wherein the first determination result is used to characterize the result of determining whether the belt conveyor is faulty based on the target model.

[0018] According to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to perform any of the belt conveyor fault determination methods.

[0019] According to another aspect of this application, a belt conveyor system is provided, comprising: a belt conveyor, an ultrasonic transmitting probe, an ultrasonic receiving probe, one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including a method for performing any of the belt conveyor fault determination methods described above.

[0020] By applying the technical solution of this application, the ELM algorithm and the VPMCD algorithm can be used together to determine whether the belt conveyor is faulty. Currently, some solutions use the VPMCD algorithm to classify the working state and fault type of rolling bearings, which can accurately identify faults and does not require a large number of samples for training. The advantage of the ELM algorithm is that it can learn a better model in a relatively short time and has high robustness to outliers, which can reduce the influence of noise in the model. This can avoid the problem that the neural network cannot converge quickly. Therefore, based on this, the target model obtained by combining the VPMCD and ELM algorithms has higher accuracy, thereby improving the detection accuracy. Attached Figure Description

[0021] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0022] Figure 1 A hardware structure block diagram of a mobile terminal for performing a method for determining belt conveyor faults according to an embodiment of this application is shown.

[0023] Figure 2 A flowchart illustrating a method for determining belt conveyor faults according to an embodiment of this application is shown.

[0024] Figure 3 A schematic diagram of the detection results using the target model is shown;

[0025] Figure 4 This diagram illustrates the results detected using a BP neural network model.

[0026] Figure 5 A schematic diagram of the corrected sample based on ELM discrimination and conventional discrimination is shown;

[0027] Figure 6 The diagram illustrates the feature weights under two different discrimination methods.

[0028] Figure 7 A schematic diagram of the weights for anomaly prediction under two different criteria is shown.

[0029] Figure 8 A structural block diagram of a belt conveyor fault determination device provided according to an embodiment of this application is shown;

[0030] Figure 9 A schematic diagram of the structure of a belt conveyor system provided according to an embodiment of this application is shown.

[0031] The above figures include the following reference numerals:

[0032] 102. Processor; 104. Memory; 106. Transmission equipment; 108. Input / output device; 100. Belt conveyor; 110. Ultrasonic transmitting probe; 120. Ultrasonic receiving probe; 130. Processor. Detailed Implementation

[0033] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0034] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0035] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0036] For ease of description, the following explains some of the nouns or terms used in the embodiments of this application:

[0037] ELM: Extreme Learning Machine (ELM) is a type of machine learning system or method based on Feedforward Neuron Network (FNN), applicable to both supervised and unsupervised learning problems.

[0038] With the development of informatization in the coal industry, dynamic target defect recognition technology for mine conveyor belts has attracted much attention and research from scholars. As research into pattern recognition technology deepens, various algorithms have emerged, including neural networks, fuzzy clustering analysis, and support vector machines. These methods have all played important roles in engineering analysis; however, they still face many shortcomings that need to be overcome. For example, support vector machines require appropriate kernel functions and parameter settings to obtain classification results that meet accuracy requirements, while artificial neural networks struggle with rapid convergence and are prone to local minima, requiring a large number of samples during training. Furthermore, these methods analyze their respective features and classification methods independently, without analyzing the differences in classification results caused by the correlation between different features.

[0039] As described in the background section, the accuracy of fault detection in existing belt conveyors is relatively low. To address the above-mentioned problems, embodiments of this application provide a method, apparatus, computer-readable storage medium, and belt conveyor system for determining belt conveyor faults.

[0040] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0041] The methods and embodiments provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal for a method of determining belt conveyor faults according to an embodiment of the present invention. Figure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0042] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the device information display method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the aforementioned networks may include wireless networks provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0043] This embodiment provides a method for determining belt conveyor faults that runs on a mobile terminal, computer terminal, or similar computing device. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0044] Figure 2 This is a flowchart illustrating a method for determining a belt conveyor fault according to an embodiment of this application. Figure 2 As shown, the method includes the following steps:

[0045] Step S201: Acquire ultrasonic data, wherein the ultrasonic data is the data of ultrasonic waves reflected by the belt conveyor after ultrasonic waves are emitted to the belt conveyor.

[0046] Specifically, this solution uses ultrasonic data to detect the belt conveyor. This ultrasonic data is obtained by emitting ultrasonic waves into the belt conveyor and then recording the ultrasonic waves reflected back from the conveyor. By analyzing and processing these reflected ultrasonic waves, relevant information about the belt conveyor can be obtained, such as its status, operating conditions, and potential faults. This technology is commonly used in industrial fields to monitor and maintain the operating condition of belt conveyors.

[0047] Step S202: Construct a target model, wherein the target model is obtained by training multiple sets of training data using the ELM algorithm and the VPMCD algorithm. Each set of training data includes historical ultrasonic data acquired within a historical time period and a first historical determination result corresponding to the historical ultrasonic data. The first historical determination result is used to characterize whether the historical belt conveyor was faulty.

[0048] Specifically, the target model can be trained using the ELM and VPMCD algorithms. The ELM-VPMCD method improves the recognition accuracy by 5% compared to the VPMCD method. The ELM-VPMCD method has better overall performance and can accurately classify samples that other methods cannot recognize, making it more accurate in identifying moving targets on mining conveyor belts.

[0049] Step S203: Input the ultrasonic data into the target model to obtain the first determination result corresponding to the ultrasonic data, wherein the first determination result is used to characterize the result of determining whether the belt conveyor is faulty based on the target model.

[0050] Specifically, because the target model has high accuracy, inputting the currently acquired ultrasonic data into the target model can yield a highly accurate first determination result, which can more accurately determine whether the conveyor belt based on the currently acquired ultrasonic data is faulty.

[0051] In this embodiment, the ELM algorithm and the VPMCD algorithm can be used together to determine whether the belt conveyor is faulty. Currently, some solutions use the VPMCD algorithm to classify the working state and fault type of rolling bearings, which can accurately identify faults and does not require a large number of samples for training. The advantage of the ELM algorithm is that it can learn a better model in a relatively short time and has high robustness to outliers, which can reduce the influence of noise in the model. This can avoid the problem that the neural network cannot converge quickly. Therefore, based on this, the target model obtained by combining the VPMCD and ELM algorithms has higher accuracy, thereby improving the detection accuracy.

[0052] Currently, some solutions have developed a Variable Prediction Model Pattern Recognition (VPMCD) method. This method, compared to other methods, primarily involves in-depth analysis of the inherent relationships between different features, and then constructing a prediction model based on these relationships. VPMCD involves two aspects: constructing the prediction model and making predictions. Feature relationships can be fitted using single or quadratic linear models and single or quadratic interactive models. While these models meet the requirements for analyzing simple working conditions, they cannot accurately fit complex relationships between feature variables for complex working conditions. To address this, many developers have conducted optimization analyses of VPMCD. For example, using an RBF neural network model to fit feature parameters, and then setting the variable prediction model using radial basis functions before pattern recognition, significantly improves the accuracy of feature value judgment.

[0053] Fault detection of mining conveyor belts can be analyzed using VPMCD. VPMCD designs a prediction model based on the intrinsic relationship between feature variables. Therefore, different feature vectors will cause significant differences in the VPMCD classifier. When one feature variable lacks a good correlation with other feature variables, using the VPMCD model to predict this feature parameter will cause a large deviation. This makes the prediction of outliers significantly affect the minimum sum of squared errors discriminant function, and may even produce incorrect results. In the specific implementation process, before building the target model, the above method also includes the following steps: building an initial model, wherein the initial model is trained using multiple sets of training data through the VPMCD algorithm. Each set of training data includes the historical ultrasonic data obtained within the above historical time period and the second historical determination result corresponding to the historical ultrasonic data. The second historical determination result is used to characterize whether the historical conveyor belt is faulty. The initial model is optimized using the ELM algorithm to obtain the target model.

[0054] In this scheme, after building the initial model with VPMCD, the ELM algorithm can be used to optimize the initial model. The target model is built through the Extreme Learning Machine. The target model can be simply referred to as the ELM-VPMCD model. In this way, only a small number of adjustable parameters need to be added to achieve a fast convergence effect, thereby further ensuring that the target model has higher accuracy.

[0055] Specifically, this scheme employs the ELM-VPMCD method, which uses Extreme Learning Machine (ELM) to optimize the initial model pattern recognition (VPMCD), to identify ultrasonic defect signals in conveyor belts. This achieves higher overall performance and results in more accurate and stable fault identification.

[0056] Various system behaviors can be quantitatively analyzed based on test characteristics and their interaction processes. Using VPMCD to establish the mathematical expression of the variable prediction model, the relationships between features can typically be defined using regression models. The formulas for linear and linear crossover are as follows:

[0057] Among them, X i Let b represent the dependent variable (target variable), b0 represent the coefficients of the model (weights or regression coefficients), and X represent the model coefficients. j The coefficient represents the independent variable (characteristic variable). This formula represents the linear relationship between the independent variable and the dependent variable. By adjusting the size of the coefficient, the data is fitted and the value of the dependent variable is predicted.

[0058] in, The error term, also known as ε, represents the linear influence of the independent variable on the dependent variable. The regression coefficient is estimated by minimizing the error term.

[0059] The model coefficients are determined based on g classes of samples, resulting in (g*p) regression models VPMCD. These regression models are then used to predict sample features. In this testing process, the squared error and a discriminant function are used to determine the type of the prediction results. For each category, the sum of squared feature prediction errors is calculated, and the result that minimizes this sum is selected as the final type.

[0060] In practical applications, the acquired signals contain numerous interference factors, making it difficult to extract complex feature values. The prediction models established using equations (1) and (2) cannot accurately reflect the relationships between the parameters. In this case, choosing an ELM network allows for fitting analysis of various complex situations. When establishing the ELM-VPMCD method using an ELM network, the VPMCD prediction model is designed using ELM functions and mathematical models. The specific steps are as follows:

[0061] Step 1: The k-classification problem contains N training samples. Each sample contains P feature values, resulting in the j-th sample xj.

[0062] Step 2: Set both k and i to 1. k is the i-th feature of the sample, and its value is k = 1, 2, ..., p).

[0063] Step 3: Train on nk k-th class samples and use the ELM network to construct VPM for the i-th feature value.

[0064] Step 4: Reassign i = i + 1, and repeat the calculation from step 3 until i = p, at which point the loop is complete.

[0065] Step 5: Assign values ​​according to the formula k = k + 1, looping between steps 3 and 5 until k reaches the value k, completing the loop. A total of P*k VPM models are trained, as shown below:

[0066]

[0067] To achieve fitting analysis of complex relationships, this scheme constructs relationship models with different features through an ELM network, and then uses VPMCD as the intrinsic model of the ELM to establish a prediction model (target model).

[0068] When performing ultrasonic testing on chemical product storage tanks or oil pipelines, the actual test objects include surface defects caused by various factors. A Q235 sheet metal sample with a thickness of 11 mm was processed, and surface defects were manually set as the ultrasonic testing area. Three surface defects of different shapes—circular, square, and elliptical—were set, with depths of 2 mm, 5 mm, and 8 mm respectively, and the interval between the different defects controlled to be 8–15 mm. A 2.2 MHz ultrasonic probe was used as the detection component, with an outer diameter of 20 mm, and the detection area was required to be larger than the area of ​​the defects in this experiment.

[0069] In the field of machine learning, it is often necessary to learn a function model using training data in order to make predictions or classifications on unknown data. Traditional neural network models require a large amount of parameter tuning and iterative optimization, resulting in low learning efficiency. Extreme Learning Machine (ELM) is a single-layer feedforward neural network based on randomization, which has become a research hotspot in recent years due to its many advantages such as fast learning speed, high accuracy, and ease of implementation.

[0070] ELM consists of three main parts:

[0071] Input layer: Responsible for receiving input sample vector X and outputting sample features.

[0072] Hidden layer: Contains K hidden nodes, the output of each hidden node is a weighted sum, i.e.

[0073]

[0074] Among them, w ik b represents the connection weight from the iii-th neuron in the input layer to the k-th neuron in the hidden layer. k It is a bias term.

[0075] Output layer: The number of neurons in the output layer equals the number of sample classes. For a binary classification problem, the output layer has two neurons, representing the positive and negative classes respectively, and the output is a vector y = (y1, y2). T .

[0076] To further improve the accuracy of the target model used to detect belt conveyor malfunctions, this application employs the ELM algorithm to optimize the initial model. This can be achieved through the following steps: obtaining a training set, wherein the training set is T = (x1, y1), (x2, y2), ..., (x n ,y n The input samples are mapped to the target feature space through the objective function to obtain the first matrix. The input samples are X = x1, x2, ..., xn. n The first matrix mentioned above is H = [h ij ] K×n Where K×n represents the dimension of the matrix, K represents the number of rows in the matrix, n represents the number of columns in the matrix, and h ij =g(z) k ), g(z) k ) represents the activation function of the k-th neuron in the hidden layer. The objective function is either the sigmoid function or the ReLU function. The connection weights between the hidden layer and the output layer are randomly initialized and converted into matrix form to obtain the output data of the output layer. Here, the connection weights are W. K×m , m is the number of neurons in the output layer, and the output data of the output layer is Y = [y i ] m×n , Y = W·H.

[0077] In this scheme, the core idea of ​​ELM is to randomly initialize the connection weights and biases between the hidden layer and the output layer, thereby achieving fast training and obtaining a high-precision model. Specifically, for a training set, the input samples are first mapped to a high-dimensional feature space through the objective function to obtain a new first matrix. Then, the connection weights between the hidden layer and the output layer are randomly initialized and represented by a matrix to obtain the output data of the output layer. This increases the diversity of the model, avoids getting trapped in local optima, and further improves the generalization ability of the algorithm.

[0078] Specifically, the n mentioned above is the same as the length n of the previous samples and the length n of the training set.

[0079] Specifically, representative and relevant features can be selected, and redundant features can be removed to improve the efficiency and accuracy of the algorithm, reduce computational costs and the risk of overfitting. Through nonlinear transformations of the hidden layers, the input data is mapped to a high-dimensional space, extracting more abstract and useful features. This allows for better extraction of abstract features from the data, improving the algorithm's expressive and predictive capabilities.

[0080] To further improve the accuracy of the target model used for detecting belt conveyor malfunctions, this application employs the ELM algorithm to optimize the initial model. This can be achieved through the following steps: calculating the fitting error between the output layer and the training data using the least squares method; obtaining the residual sum of squares, where the residual sum of squares is...

[0081]

[0082] y i Let t represent an n-dimensional vector. i Represent an n-dimensional vector; convert the above residual sum of squares into a second matrix, where the above second matrix is...

[0083]

[0084] Where, T = [t ij ] m×n T represents the target vector of the output layer, and the weights of the output layer are W = (H T H) - 1 H T H.

[0085] In this scheme, the least squares method can be used to calculate the fitting error between the output layer and the training data, and the sum of squared residuals can be calculated. Then, the matrix is ​​rewritten to obtain the second matrix. For a fixed number of neurons in the hidden layer, it can also be calculated using W = (H... T H) -1 H T H is calculated, therefore, the learning process of ELM can be completed by only finding the weights from the input layer to the hidden layer and the weights from the hidden layer to the output layer. This makes the output of the target model as close as possible to the true value, further improving the accuracy and stability of the algorithm.

[0086] Specifically, the ELM algorithm optimizes the initial model. In ELM, the fitting error and the sum of squared residuals are used to optimize the parameters of the initial model. First, the ELM algorithm initializes the parameters of the initial model by randomly generating the weight matrix from the input layer to the hidden layer and the bias vector of the hidden layer. Then, the ELM algorithm uses these parameters to calculate the output of the hidden layer. The output of the hidden layer can be calculated using the following formula: H = G(X*W+b), where X is the input data matrix, W is the weight matrix, b is the bias vector, and G is the activation function. Next, the ELM algorithm uses the least squares method to fit the weight matrix of the output layer. The goal of the least squares method is to minimize the fitting error, that is, to minimize the difference between the predicted value and the true value of the output layer. The solution of the least squares method can be obtained using the following formula: O = H*beta, where O is the predicted value of the output layer, and beta is the weight matrix of the output layer. Then, the ELM algorithm calculates the fitting error and the sum of squared residuals. The fitting error can be obtained by calculating the difference between the predicted value and the true value. The sum of squared residuals can be obtained by calculating the sum of squared fitting errors. Finally, the ELM algorithm can optimize the initial model by adjusting the parameters of the hidden layers (weight matrix and bias vector). Optimization algorithms such as gradient descent can be used to minimize the fitting error and the sum of squared residuals, resulting in a better initial model.

[0087] y i It is the result of the algorithm output, t i It is the true value corresponding to the training sequence, which is the target value, y. i and t i The difference (residual) reflects the goodness of fit; the smaller the residual, the better.

[0088] To further improve the accuracy of the target model used to detect whether a belt conveyor is malfunctioning, this application uses the ELM algorithm to optimize the initial model, which can be achieved through the following steps: obtaining the bias function between the hidden layer and the output layer; randomly initializing the bias function and optimizing the bias function using a target optimization algorithm to obtain the optimized bias function, wherein the target optimization algorithm is a learning rate decay algorithm or an adaptive learning rate algorithm.

[0089] In this scheme, the bias function can be optimized. Optimizing the bias function can make the model better adapt to the training data, improve the model's fitting ability, accelerate the model's convergence speed, reduce training time, reduce the risk of overfitting, improve the model's generalization ability, make the model more stable, reduce the model's differences under different initialization conditions, and improve the model's reliability.

[0090] Specifically, optimizing the bias function can be performed through the following steps: Step 1: First, randomly initialize the biases of the hidden layer and the output layer. Step 2: Using the randomly initialized biases, pass the input data to the output layer through a nonlinear transformation of the hidden layer and calculate the output value. Step 3: Compare the output value with the true value, calculate the loss function, and measure the model's prediction error. Step 4: Use an optimization algorithm (such as gradient descent, Adam, etc.) to update the biases of the hidden layer and the output layer based on the gradient information of the loss function, so that the loss function is minimized. Step 5: Repeat steps 2-4 until the predetermined stopping conditions are met (such as reaching the maximum number of iterations, the loss function converging, etc.).

[0091] In some embodiments, after optimizing the initial model using the ELM algorithm, the method further includes the following step: obtaining the sample x to be determined using the following formula:

[0092]

[0093] The samples to be determined are mapped to hidden layers through an activation function; the product of the output of the mapped hidden layer and the connection weights from the hidden layer to the output layer is calculated to obtain the calculation result, where the calculation result is y = W·h; the calculation result is then classified to determine the category of the samples to be determined.

[0094] In this scheme, after obtaining the weights of the target model, new samples can be used to calculate the results of the output layer. For a sample to be determined, it can be mapped to the hidden layer through the activation function, and then the calculation result can be obtained. Finally, the category of the sample can be obtained by classifying the calculation result. Since the target model has been optimized, the category here is relatively accurate and has high precision.

[0095] To verify the recognition effect of this solution, tests can be conducted to verify the accuracy of the detection effect. In some embodiments, after inputting the ultrasonic data into the target model and obtaining the determination result corresponding to the ultrasonic data, the method further includes the following steps: obtaining a reference determination result, wherein the reference determination result is the result of fault identification of the belt conveyor standard; inputting the ultrasonic data into the initial model to obtain a second determination result corresponding to the ultrasonic data, wherein the second determination result is used to characterize the result of determining whether the belt conveyor is faulty based on the initial model; calculating a first error value between the first determination result and the reference determination result, and calculating a second error value between the second determination result and the reference determination result; displaying the first error value and the second error value in a target display format on the display interface, wherein the target display format is a chart display and / or a table display.

[0096] In this scheme, the prediction results of ELM-VPMCD can be compared with the reference results to verify the accuracy of the ELM-VPMCD model in this scheme, and thus it can be seen that the accuracy of the ELM-VPMCD model is high.

[0097] To verify the beneficial properties of the ELM discriminant function, a comparison was conducted between the ELM-VPMCD method and the BP neural network. The test results are shown below. Figure 3 and Figure 4 In the vertical axis, 1, 2, 3, and 4 represent defects of 40mm², 80mm², 120mm², and 160mm², respectively.

[0098] from Figure 3 and Figure 4 As can be seen, since VPMCD is initialized in a random manner, the test results fluctuate greatly, with the actual fluctuation range between 82% and 89%. In this test, the accuracy of the results was calculated 10 times and then the 85% index was determined as the optimal result of the BP neural network.

[0099] Comparative test results show that the ELM-VPMCD method improves the recognition accuracy by 5% compared to the VPMCD method. The ELM-VPMCD method has superior overall performance and can accurately classify samples that other methods cannot recognize.

[0100] The sample classification result should belong to defect type 2, with a corresponding area of ​​80 mm². The prediction error squared matrix is ​​processed using both conventional and ELM discriminant functions. According to Table 1, when judging the test sample using the conventional method, the result belongs to "defect type 4".

[0101] Table 1

[0102]

[0103]

[0104] Ideally, all features with minimum squared errors are included in "Defect Type 2". 31 features in the sample have minimum errors belonging to "Defect Type 2", demonstrating relatively ideal predictive performance. However, processing with conventional discrimination methods can lead to errors in the classification results. This is because some predicted anomalous features have excessive weight in conventional methods, resulting in incorrect classification. Table 2 shows the results after the prediction error transformation, revealing a normalized matrix with consistent dimensions. Accurately classifying the test samples as "Defect Type 2" meets the actual judgment requirements.

[0105] Table 2

[0106]

[0107]

[0108] A total of 100 test samples were selected as test subjects, of which 12 samples produced results consistent with the test samples. Figure 5 As shown, ELM-VPMCD can accurately classify samples that are incorrectly identified by conventional methods.

[0109] This scheme can also test the performance of different methods in judging samples; the results are shown in [link to results]. Figure 6 When the calculated weight is closer to 1, this feature will have a significant impact on the judgment result.

[0110] Experimental tests showed that all three selected features produced incorrect predictions and generated large squared error values. Figure 6 As shown, when judging using conventional methods, the feature weights obtained from the test are 33%, 46%, and 1% respectively, with a total weight ratio of 90%; the weights of the aforementioned features are 17%, 19%, and 10% respectively, with a total weight of less than 50%. For anomaly prediction features, the discrimination weight is reduced by 47%. For accurately predictive features (typically with an error within 1), the weight is amplified, increasing the cumulative weight to 51%.

[0111] from Figure 7 The results of testing anomaly prediction feature weights on similar test samples can be seen. It is evident that the ELM-VPMCD method can significantly weaken the prediction of anomalous features and amplify the proportion of features with smaller prediction errors, thereby ensuring that the classification results are more accurate and stable than conventional methods.

[0112] Therefore, the main advantage of this solution is:

[0113] 1) The ELM-VPMCD method improves the recognition accuracy by 5% compared to the VPMCD method. The ELM-VPMCD method has better overall performance and can accurately classify samples that cannot be recognized by other methods.

[0114] 2) The ELM discriminant method produces results consistent with the test sample's feature value error distribution. When processed with ELM-VPMCD, it can accurately classify samples that are incorrectly identified by conventional methods.

[0115] 3) The ELM-VPMCD method can significantly weaken the predicted anomalous features and amplify the proportion of features with small prediction errors, ensuring that the classification results are more accurate and more stable than conventional methods.

[0116] This application also provides a device for determining belt conveyor faults. It should be noted that this device can be used to execute the belt conveyor fault determination method provided in this application. This device is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0117] The following describes the belt conveyor fault determination device provided in the embodiments of this application.

[0118] Figure 8 This is a structural block diagram of a belt conveyor fault determination device according to an embodiment of this application. Figure 8 As shown, the device includes:

[0119] The first acquisition unit 10 is used to acquire ultrasonic data, wherein the ultrasonic data is the data of ultrasonic waves reflected by the belt conveyor after ultrasonic waves are emitted to the belt conveyor.

[0120] The first construction unit 20 is used to construct a target model, wherein the target model is obtained by training multiple sets of training data using the ELM algorithm and the VPMCD algorithm. Each set of training data includes historical ultrasonic data acquired within a historical time period and a first historical determination result corresponding to the historical ultrasonic data. The first historical determination result is used to characterize whether the historical belt conveyor was faulty.

[0121] The determining unit 30 is used to input the ultrasonic data into the target model to obtain a first determining result corresponding to the ultrasonic data, wherein the first determining result is used to characterize the result of determining whether the belt conveyor is faulty based on the target model.

[0122] In this embodiment, the ELM algorithm and the VPMCD algorithm can be used together to determine whether the belt conveyor is faulty. Currently, some solutions use the VPMCD algorithm to classify the working state and fault type of rolling bearings, which can accurately identify faults and does not require a large number of samples for training. The advantage of the ELM algorithm is that it can learn a better model in a relatively short time and has high robustness to outliers, which can reduce the influence of noise in the model. This can avoid the problem that the neural network cannot converge quickly. Therefore, based on this, the target model obtained by combining the VPMCD and ELM algorithms has higher accuracy, thereby improving the detection accuracy.

[0123] Fault detection of mining conveyor belts can be analyzed using VPMCD. VPMCD designs a prediction model based on the intrinsic relationship between feature variables. Therefore, different feature vectors will cause significant differences in the VPMCD classifier. When one feature variable lacks a good correlation with other feature variables, using the VPMCD model to predict this feature parameter will cause a large deviation. This makes the prediction of outliers significantly affect the minimum sum of squared errors discriminant function, and may even produce incorrect results. In the specific implementation process, the above device also includes a second construction unit and an optimization unit. The second construction unit is used to build an initial model before building the target model. The initial model is trained using multiple sets of training data through the VPMCD algorithm. Each set of training data includes the historical ultrasonic data obtained within the historical time period and the second historical determination result corresponding to the historical ultrasonic data. The second historical determination result is used to characterize whether the historical conveyor belt is faulty. The optimization unit is used to optimize the initial model using the ELM algorithm to obtain the target model.

[0124] In this scheme, after building the initial model with VPMCD, the ELM algorithm can be used to optimize the initial model. The target model is built through the Extreme Learning Machine. The target model can be simply referred to as the ELM-VPMCD model. In this way, only a small number of adjustable parameters need to be added to achieve a fast convergence effect, thereby further ensuring that the target model has higher accuracy.

[0125] To further improve the accuracy of the target model used to detect belt conveyor malfunctions, the optimization unit of this application includes a first acquisition module and a first processing module. The first acquisition module is used to acquire a training set, wherein the training set is T = (x1, y1), (x2, y2), ..., (x n ,y n The input samples are mapped to the target feature space through the objective function to obtain the first matrix. The input samples are X = x1, x2, ..., xn. n The first matrix mentioned above is H = [h ij ] K×n Where K×n represents the dimension of the matrix, K represents the number of rows in the matrix, n represents the number of columns in the matrix, and h ij =g(z) k ), g(z) k ) represents the activation function of the k-th neuron in the hidden layer, and the objective function is either the sigmoid function or the ReLU function; the first processing module is used to randomly initialize the connection weights between the hidden layer and the output layer, and convert them into matrix form to obtain the output data of the output layer, where the connection weights are W. K×m, m is the number of neurons in the output layer, and the output data of the output layer is Y = [y i ] m×n , Y = W·H.

[0126] In this scheme, the core idea of ​​ELM is to randomly initialize the connection weights and biases between the hidden layer and the output layer, thereby achieving fast training and obtaining a high-precision model. Specifically, for a training set, the input samples are first mapped to a high-dimensional feature space through the objective function to obtain a new first matrix. Then, the connection weights between the hidden layer and the output layer are randomly initialized and represented by a matrix to obtain the output data of the output layer. This increases the diversity of the model, avoids getting trapped in local optima, and further improves the generalization ability of the algorithm.

[0127] To further improve the accuracy of the target model used to detect belt conveyor malfunctions, the optimization unit of this application includes a calculation module, a second acquisition module, and a second processing module. The calculation module is used to calculate the fitting error between the output layer and the training data using the least squares method; the second acquisition module is used to acquire the residual sum of squares, wherein the residual sum of squares is...

[0128]

[0129] y i Let t represent an n-dimensional vector. i Representing an n-dimensional vector; the second processing module is used to convert the above residual sum of squares into a second matrix, wherein the above second matrix is...

[0130]

[0131] Where, T = [t ij ] m×n T represents the target vector of the output layer, and the weights of the output layer are W = (H T H) - 1 H T H.

[0132] In this scheme, the least squares method can be used to calculate the fitting error between the output layer and the training data, and the sum of squared residuals can be calculated. Then, the matrix is ​​rewritten to obtain the second matrix. For a fixed number of neurons in the hidden layer, it can also be calculated using W = (H... T H) -1 H T H is calculated, therefore, the learning process of ELM can be completed by only finding the weights from the input layer to the hidden layer and the weights from the hidden layer to the output layer. This makes the output of the target model as close as possible to the true value, further improving the accuracy and stability of the algorithm.

[0133] To further improve the accuracy of the target model used to detect whether a belt conveyor is malfunctioning, the optimization unit of this application includes a third acquisition module and a third processing module. The third acquisition module is used to acquire the bias function between the hidden layer and the output layer. The third processing module is used to randomly initialize the above bias function and optimize the above bias function using a target optimization algorithm to obtain the optimized bias function. The above target optimization algorithm is a learning rate decay algorithm or an adaptive learning rate algorithm.

[0134] In this scheme, the bias function can be optimized. Optimizing the bias function can make the model better adapt to the training data, improve the model's fitting ability, accelerate the model's convergence speed, reduce training time, reduce the risk of overfitting, improve the model's generalization ability, make the model more stable, reduce the model's differences under different initialization conditions, and improve the model's reliability.

[0135] In some embodiments, the above-described apparatus further includes a second acquisition unit, a first calculation unit, and a classification unit. The second acquisition unit is used to acquire the sample x to be determined after optimizing the initial model using the ELM algorithm, by means of the following formula:

[0136]

[0137] The above-mentioned samples to be determined are mapped to the hidden layer through an activation function; the first calculation unit is used to calculate the product of the output of the mapped hidden layer and the connection weight from the hidden layer to the output layer to obtain the calculation result, wherein the calculation result is y = W·h; the classification unit is used to classify the calculation result to determine the category of the above-mentioned samples to be determined.

[0138] In this scheme, after obtaining the weights of the target model, new samples can be used to calculate the results of the output layer. For a sample to be determined, it can be mapped to the hidden layer through the activation function, and then the calculation result can be obtained. Finally, the category of the sample can be obtained by classifying the calculation result. Since the target model has been optimized, the category here is relatively accurate and has high precision.

[0139] To verify the recognition effect of this solution, tests can be conducted to verify the accuracy of the detection effect. In some embodiments, the above-mentioned device further includes a third acquisition unit, a first processing unit, a second calculation unit, and a second processing unit. The third acquisition unit is used to acquire a reference determination result after inputting the ultrasonic data into the target model and obtaining the determination result corresponding to the ultrasonic data, wherein the reference determination result is the result of fault identification of the belt conveyor standard. The first processing unit is used to input the ultrasonic data into the initial model and obtain a second determination result corresponding to the ultrasonic data, wherein the second determination result is used to characterize the result of determining whether the belt conveyor is faulty based on the initial model. The second calculation unit is used to calculate a first error value between the first determination result and the reference determination result, and calculate a second error value between the second determination result and the reference determination result. The second processing unit is used to display the first error value and the second error value in a target display form on the display interface, wherein the target display form is a chart display and / or a table display.

[0140] In this scheme, the prediction results of ELM-VPMCD can be compared with the reference results to verify the accuracy of the ELM-VPMCD model in this scheme, and thus it can be seen that the accuracy of the ELM-VPMCD model is high.

[0141] The aforementioned belt conveyor fault determination device includes a processor and a memory. The first acquisition unit, the first construction unit, and the determination unit are all stored as program units in the memory. The processor executes these program units stored in the memory to achieve the corresponding functions. All of the above modules are located in the same processor; alternatively, the modules may be located in different processors in any combination.

[0142] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and adjusting kernel parameters can address the low accuracy of fault detection in existing belt conveyor technologies.

[0143] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0144] This invention provides a computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the method for determining belt conveyor faults.

[0145] This invention provides a processor for running a program, wherein the program executes the method for determining belt conveyor faults.

[0146] This application also provides a belt conveyor system, such as Figure 9 As shown, the device includes a belt conveyor 100, an ultrasonic transmitting probe 110, an ultrasonic receiving probe 120, one or more processors 130, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for performing any of the above-described belt conveyor fault determination methods.

[0147] Specifically, the ultrasonic transmitting probe can be a digital ultrasonic transmitting probe, and the ultrasonic receiving probe can be a digital ultrasonic receiving probe.

[0148] Specifically, the belt conveyor is used to transport targets including, but not limited to, various types of coal. A digital ultrasonic transmitting probe is used to transmit ultrasonic waves at frequencies including, but not limited to, 100kHz, to the mining conveyor equipment. An ultrasonic receiving probe is used to receive ultrasonic waves reflected from the mining conveyor equipment. The ultrasonic detector server includes one or more processors that process the ultrasonic information received by the digital ultrasonic receiving probe using an ELM-VPMCD model.

[0149] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of a method for determining belt conveyor faults. The device described herein can be a server, PC, tablet, mobile phone, etc.

[0150] This application also provides a computer program product that, when executed on a data processing device, is adapted to perform the initialization method steps for determining at least one belt conveyor fault.

[0151] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0152] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0153] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0154] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0155] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0156] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0157] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0158] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0159] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0160] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0161] 1) The belt conveyor fault determination method of this application can use the ELM algorithm and the VPMCD algorithm together to determine whether the belt conveyor is faulty. Currently, some solutions use the VPMCD algorithm to classify the working state and fault type of rolling bearings, which can accurately determine the fault and does not require a large number of samples for training. The advantage of the ELM algorithm is that it can learn a better model in a relatively short time and has high robustness to outliers, which can reduce the influence of noise in the model. This can avoid the problem that the neural network cannot converge quickly. Therefore, on this basis, the target model obtained by combining the VPMCD and ELM algorithms has higher accuracy, thereby improving the detection accuracy.

[0162] 2) The belt conveyor fault determination device of this application can use the ELM algorithm and the VPMCD algorithm together to determine whether the belt conveyor is faulty. Currently, some solutions use the VPMCD algorithm to classify the working state and fault type of rolling bearings, which can accurately determine the fault and does not require a large number of samples for training. The advantage of the ELM algorithm is that it can learn a better model in a relatively short time and has high robustness to outliers, which can reduce the influence of noise in the model. This can avoid the problem that the neural network cannot converge quickly. Therefore, on this basis, the target model obtained by combining the VPMCD and ELM algorithms has higher accuracy, thereby improving the detection accuracy.

[0163] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for determining belt conveyor faults, characterized in that, include: Acquire ultrasonic data, wherein the ultrasonic data is the data of ultrasonic waves reflected by the belt conveyor after ultrasonic waves are emitted to the belt conveyor; An initial model is constructed, wherein the initial model is trained using multiple sets of training data through the VPMCD algorithm. Each set of training data includes historical ultrasonic data acquired within a historical time period and a second historical determination result corresponding to the historical ultrasonic data. The second historical determination result is used to characterize whether the belt conveyor was faulty in the past. The initial model is optimized using the ELM algorithm to obtain a target model. The ultrasonic data is input into the target model to obtain a first determination result corresponding to the ultrasonic data. The first determination result is used to characterize whether the belt conveyor was faulty based on the target model. The initial model is optimized using the ELM algorithm, including: obtaining a training set, wherein the training set is... The input samples are mapped to the target feature space through the objective function to obtain the first matrix, where the input samples are... The first matrix is ,in, Represents the dimension of the matrix. Represents the rows of a matrix. Represents the columns of a matrix. , Indicates the hidden layer number 1 The activation function of each neuron is determined, with the objective function being either the sigmoid function or the ReLU function. The connection weights between the hidden layer and the output layer are randomly initialized and converted into matrix form to obtain the output data of the output layer. The connection weights are... , The number of neurons in the output layer is [number], and the output data of the output layer is [data]. ; The initial model is optimized using the ELM algorithm, including: calculating the fitting error between the output layer and the training data using the least squares method; and obtaining the residual sum of squares, wherein the residual sum of squares is... , Represents an n-dimensional vector. Represent an n-dimensional vector; convert the sum of squared residuals into a second matrix, where the second matrix is... ,in, , Let the target vector of the output layer be and the weights of the output layer be . ; Among these, various system behaviors are quantitatively analyzed based on test characteristics and their interaction processes. A variable prediction model mathematical expression is established using VPMCD. Feature relationships are defined using a regression model, and linear and linear cross formulas are provided. ;in, Indicates the dependent variable. Represents the coefficients of the model. The independent variable is represented by the coefficient, and this formula represents the linear relationship between the independent variable and the dependent variable. The data is fitted and the value of the dependent variable is predicted by adjusting the size of the coefficient. ;in, This refers to the error term, also simply called... This formula represents the linear influence of the independent variable on the dependent variable, estimating the regression coefficient by minimizing the error term. The first step involves N training samples in a k-class classification problem, each containing P feature values, resulting in the j-th sample xj. The second step sets both k and i to 1, where k is the i-th feature of the sample, with values ​​k = 1, 2, ..., p. The third step trains on nk k-th class samples, constructing a VPM for the i-th feature value using an ELM network. The fourth step reassigns i = i + 1, iterating through the third step until i = p. The fifth step assigns values ​​according to k = k + 1, iterating between the third and fifth steps until k reaches the value k. A total of P samples are trained. k VPM models, the results are We construct relational models with different features using ELM networks, and then use VPMCD as the intrinsic model of ELM to build a prediction model.

2. The method according to claim 1, characterized in that, The initial model is optimized using the ELM algorithm, including: Obtain the bias function between the hidden layer and the output layer; The bias function is randomly initialized and then optimized using a target optimization algorithm to obtain the optimized bias function. The target optimization algorithm is either a learning rate decay algorithm or an adaptive learning rate algorithm.

3. The method according to claim 1, characterized in that, After inputting the ultrasonic data into the target model and obtaining the determination result corresponding to the ultrasonic data, the method further includes: Obtain a reference determination result, wherein the reference determination result is the fault identification result of the belt conveyor standard; The ultrasonic data is input into the initial model to obtain a second determination result corresponding to the ultrasonic data, wherein the second determination result is used to characterize the result of determining whether the belt conveyor is faulty based on the initial model; Calculate a first error value between the first determination result and the reference determination result, and calculate a second error value between the second determination result and the reference determination result; The first error value and the second error value are displayed in the display interface in a target display format, wherein the target display format is a chart display and / or a table display.

4. A device for determining belt conveyor faults, characterized in that, include: The first acquisition unit is used to acquire ultrasonic data, wherein the ultrasonic data is the data of ultrasonic waves reflected by the belt conveyor after ultrasonic waves are emitted to the belt conveyor; The second construction unit is used to construct an initial model before constructing the target model. The initial model is obtained by training multiple sets of training data using the VPMCD algorithm. Each set of training data includes historical ultrasonic data acquired within a historical time period and a second historical determination result corresponding to the historical ultrasonic data. The second historical determination result is used to characterize whether the belt conveyor was faulty in the past. The optimization unit is used to optimize the initial model using the ELM algorithm to obtain the target model. A determining unit is used to input the ultrasonic data into the target model to obtain a first determining result corresponding to the ultrasonic data, wherein the first determining result is used to characterize the result of determining whether the belt conveyor is faulty based on the target model; The device further includes: The optimization unit includes a first acquisition module and a first processing module. The first acquisition module is used to acquire a training set, wherein the training set is... The input samples are mapped to the target feature space through the objective function to obtain the first matrix, where the input samples are... The first matrix is ,in, Represents the dimension of the matrix. Represents the rows of a matrix. Represents the columns of a matrix. , Indicates the hidden layer number 1 The activation function of each neuron is specified, and the objective function is either the sigmoid function or the ReLU function. The first processing module is used to randomly initialize the connection weights between the hidden layer and the output layer, and convert them into matrix form to obtain the output data of the output layer, wherein the connection weights are... , The number of neurons in the output layer is [number], and the output data of the output layer is [data]. ; The optimization unit includes a calculation module, a second acquisition module, and a second processing module. The calculation module is used to calculate the fitting error between the output layer and the training data using the least squares method. The second acquisition module is used to acquire the residual sum of squares, wherein the residual sum of squares is... , Represents an n-dimensional vector. Representing an n-dimensional vector; the second processing module is used to convert the residual sum of squares into a second matrix, wherein the second matrix is... ,in, , Let the target vector of the output layer be and the weights of the output layer be . ; Among these, various system behaviors are quantitatively analyzed based on test characteristics and their interaction processes. A variable prediction model mathematical expression is established using VPMCD. Feature relationships are defined using a regression model, and linear and linear cross formulas are provided. ;in, Indicates the dependent variable. Represents the coefficients of the model. The independent variable is represented by the coefficient, and this formula represents the linear relationship between the independent variable and the dependent variable. The data is fitted and the value of the dependent variable is predicted by adjusting the size of the coefficient. ;in, This refers to the error term, also simply called... This formula represents the linear influence of the independent variable on the dependent variable, and estimates the regression coefficient by minimizing the error term; The device is further configured in the following steps: First, the k-classification problem contains N training samples, each containing P feature values, to obtain the j-th sample xj; Second, set both k and i to 1, where k is the i-th feature of the sample, with a value condition of k=1, 2, ..., p; Third, train on nk k-class samples, constructing a VPM using an ELM network for the i-th feature value; Fourth, reassign i=i+1, iteratively calculating the third step until i=p; Fifth, reassign values ​​according to k=k+1, iteratively processing between the third and fifth steps until k is set to k; A total of P samples are trained. k VPM models, the results are We construct relational models with different features using ELM networks, and then use VPMCD as the intrinsic model of ELM to build a prediction model.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the belt conveyor fault determination method according to any one of claims 1 to 3.

6. A belt conveyor system, characterized in that, include: The belt conveyor, an ultrasonic transmitting probe, an ultrasonic receiving probe, one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including a method for performing the belt conveyor fault determination method according to any one of claims 1 to 3.

Images