Chain transmission tension real-time monitoring method for scraper conveyor

By using Transformer model and sensor data, real-time monitoring and prediction of scraper conveyor chain tension is solved, and traditional methods are difficult to respond to tension changes, achieving more efficient equipment control and maintenance, reducing maintenance costs.

CN119085904APending Publication Date: 2024-12-06SHANDONG UNIV OF SCI & TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411100041.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Traditional methods are difficult to accurately respond to real-time changes in the chain tension of scraper conveyors and predict future pulsation trends, resulting in the inability to effectively control the operation of the equipment and affect production efficiency.

Method used

The Transformer model is used to combine sensor data, and by obtaining and integrating attribute data, operation data and tension data, a tension prediction model is built, real-time monitoring and prediction, identify abnormal data points and analyze their causes.

Benefits of technology

Real-time monitoring and prediction of scraper conveyor tension is realized, problems are discovered in a timely manner, equipment failures are avoided, maintenance costs are reduced, emergency repairs are avoided, and equipment reliability and production efficiency are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119085904A_ABST
    Figure CN119085904A_ABST
Patent Text Reader

Abstract

The invention discloses a chain transmission tension real-time monitoring method for a scraper conveyor, and belongs to the technical field of chain transmission tension monitoring for the scraper conveyor, and the method comprises the steps: obtaining the attribute data of the scraper conveyor, and obtaining the operation data and tension data of the scraper conveyor in a working state and a non-working state through a sensor; integrating the attribute data, the operation data and the tension data to obtain a to-be-calculated data set; a Transform model is constructed, the Transform model is trained based on the sample set, and a tension prediction model is obtained; inputting the data set to be calculated into the tension prediction model for calculation to obtain a tension prediction curve; and acquiring data points which do not conform to the normal operation range based on the tension prediction curve, and analyzing a tension anomaly source based on the data points which do not conform to the normal operation range. According to the method, real-time data analysis can be realized through the Transform model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of chain transmission tension monitoring for scraper conveyors, and in particular relates to a real-time monitoring method for chain transmission tension for scraper conveyors. Background Art

[0002] The fully-mechanized scraper conveyor is one of the main supporting equipment of the fully-mechanized working face. Its reliable, safe and efficient operation directly affects the production efficiency of the coal mine. The tension state of the chain can be used as an important aspect to evaluate the working condition of the scraper conveyor. The tension in the belt may also change depending on the position of the belt in the conveyor system. The tension in a part of the belt when it is located in the input area and the discharge area may be different from the tension in the same part of the belt when it is located in the spiral conveyor stack. Furthermore, the tension in the belt may change during the operation of the conveyor system, and the tension in the clean belt at the beginning of the operation may be different from the tension in the used belt with ice, products or other debris attached to it at the end of the operation.

[0003] Traditional methods focus on using mechanical adjustment devices or simple control systems to manage chain tension, but these methods often cannot accurately respond to real-time data changes, and it is difficult to effectively adapt to rapidly changing working conditions or predict future pulsation trends. Therefore, there is an urgent need to develop a more advanced and intelligent system that can monitor and accurately predict tension pulsation in real time, thereby achieving more effective control strategies. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a real-time monitoring method for chain transmission tension of a scraper conveyor to solve the problems existing in the above prior art.

[0005] To achieve the above object, the present invention provides a method for real-time monitoring of chain transmission tension for a scraper conveyor, comprising:

[0006] Acquire the attribute data of the scraper conveyor, and acquire the operation data and tension data of the scraper conveyor in the working state and the non-working state through the sensor;

[0007] Integrate the attribute data, operation data and tension data to obtain a data set to be calculated;

[0008] Constructing a Transformer model, and training the Transformer model based on the sample set to obtain a tension prediction model;

[0009] Inputting the data set to be calculated into the tension prediction model for calculation to obtain a tension prediction curve;

[0010] Data points that do not conform to a normal operating range are acquired based on the tension prediction curve, and sources of tension anomalies are analyzed based on the data points that do not conform to a normal operating range.

[0011] Preferably, the process of integrating the attribute data, operation data and tension data to obtain the data set to be calculated includes:

[0012] Correcting erroneous or missing data points in the attribute data, operation data, and tension data to obtain a corrected data set;

[0013] Synchronizing the timestamps in the corrected data set to generate aligned data;

[0014] Merging the alignment data based on the device ID of the scraper conveyor to generate a merged data set;

[0015] The merged data set is tested to obtain the data set to be calculated.

[0016] Preferably, the process of verifying the merged data set to obtain the data set to be calculated includes:

[0017] Performing a normality test on the combined data set to obtain normally distributed data and non-normally distributed data;

[0018] The normally distributed data are subjected to a t test, and the non-normally distributed data are subjected to a parameter rank sum test to obtain the data set to be calculated.

[0019] Preferably, the process of constructing the Transformer model includes:

[0020] Construct an initial Transformer model with a multi-head attention mechanism, a first residual network, a feedforward neural network, and a second residual network connected in sequence;

[0021] Processing the output sequence of the Transformer model to obtain a processing model;

[0022] The fully connected layer is used as the output layer of the processing model to generate the Transformer model.

[0023] Preferably, the expression of the multi-head attention mechanism is:

[0024] MHSA(x)=Concat(head1,…,head m )W O ;

[0025]

[0026] Among them, B relis the relative position information matrix, which is obtained by multiplying the relative position distance by a specific factor; Q, K, and V are the query, index, and content vectors of the input data, respectively. are the weight matrices corresponding to the i-th head Q, K, and V, respectively, and W O is the weight matrix, m represents the number of heads; d K Represents the feature dimension.

[0027] Preferably, the process of training the Transformer model based on the sample set includes:

[0028] Pre-training the Transformer model through the BERT pre-training model to obtain a pre-training model;

[0029] Acquire a sample set, perform iterative training on a pre-trained model based on the sample set, and terminate the iteration after the pre-trained model reaches automatic learning of input attribute data, operation data, and tension data, thereby obtaining the tension prediction model;

[0030] The loss function of the Transformer model is the cross entropy loss function.

[0031] Preferably, the process of iteratively training the pre-trained model based on the sample set further includes:

[0032] Dividing the sample set of the tension detection into a training set and a test set;

[0033] The pre-trained model is trained using the training set, a cross entropy loss function is selected, and the cross entropy loss function is minimized by a stochastic gradient descent algorithm to obtain a stochastic gradient descent model;

[0034] Using a validation set to validate the performance of the stochastic gradient descent model, to obtain a validation result;

[0035] Based on the verification result, the learning rate and the number of pooling layers of the stochastic gradient descent model are optimized to obtain a trained model.

[0036] Preferably, the process of inputting the data set to be calculated into the tension prediction model for calculation to obtain the tension prediction curve includes:

[0037] Inputting the data set to be calculated into the tension prediction model for calculation to obtain a calculation result;

[0038] The calculation results are visually interpreted to obtain the tension prediction curve.

[0039] Preferably, the process of acquiring data points that do not conform to the normal operating range based on the tension prediction curve, and analyzing the source of the tension anomaly based on the data points that do not conform to the normal operating range includes:

[0040] Identifying outliers in the data distribution of the tension prediction curve based on the isolation forest;

[0041] Analyze the causal relationship between the abnormal point and the attribute data and the operation data to obtain an analysis result;

[0042] The analysis results are compiled into a report, which includes a detailed description of the anomaly, the presumed cause, and the suggested solution.

[0043] Compared with the prior art, the present invention has the following advantages and technical effects:

[0044] The Transformer model in the present invention can realize real-time data analysis, which is particularly important for important equipment such as scraper conveyors that require continuous monitoring. Real-time monitoring can immediately detect problems, such as abnormal tension, so that measures can be taken quickly to avoid possible equipment failure or damage. And through predictive maintenance rather than responsive maintenance, the present invention allows enterprises to allocate resources more effectively and avoid unnecessary maintenance activities, thereby reducing overall maintenance costs. Real-time monitoring can also help avoid expensive emergency repairs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0046] Figure 1 The present invention is a flowchart of a method for real-time monitoring of chain transmission tension for a scraper conveyor according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0048] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0049] Embodiment 1

[0050] like Figure 1 As shown, this embodiment provides a real-time monitoring method for chain transmission tension of a scraper conveyor, comprising:

[0051] Acquire the attribute data of the scraper conveyor, and acquire the operation data and tension data of the scraper conveyor in the working state and the non-working state through the sensor;

[0052] Integrate the attribute data, operation data and tension data to obtain a data set to be calculated;

[0053] Constructing a Transformer model, and training the Transformer model based on the sample set to obtain a tension prediction model;

[0054] Inputting the data set to be calculated into the tension prediction model for calculation to obtain a tension prediction curve;

[0055] Data points that do not conform to a normal operating range are acquired based on the tension prediction curve, and sources of tension anomalies are analyzed based on the data points that do not conform to a normal operating range.

[0056] Install tension sensors at both ends of the chain or at appropriate locations to monitor the tension of the chain in real time. Sensors include load cells, pressure sensors or strain gauges. The choice of sensor depends on the size of the chain, budget and required accuracy. Connect the sensor to a data acquisition system (DAQ). The DAQ system can collect sensor data and convert this data into usable information. And collect data under different working conditions, including chain tension data under different loads, speeds and environmental conditions.

[0057] Further optimizing the scheme, integrating the attribute data, operation data and tension data, and obtaining the data set to be calculated includes:

[0058] Correcting erroneous or missing data points in the attribute data, operation data, and tension data to obtain a corrected data set;

[0059] Synchronizing the timestamps in the corrected data set to generate aligned data;

[0060] Merging the alignment data based on the device ID of the scraper conveyor to generate a merged data set;

[0061] The merged data set is tested to obtain the data set to be calculated.

[0062] The attribute data, operation data and tension data are combined into one data table. In this embodiment, this is accomplished through a common identifier, such as a device ID.

[0063] Further optimizing the scheme, testing the merged data set, and obtaining the data set to be calculated includes:

[0064] Performing a normality test on the combined data set to obtain normally distributed data and non-normally distributed data;

[0065] The normally distributed data are subjected to a t test, and the non-normally distributed data are subjected to a parameter rank sum test to obtain the data set to be calculated.

[0066] The process of t-test is:

[0067] Establishing Hypothesis: Before conducting a t-test, two hypotheses must be established: the null hypothesis (H0) and the alternative hypothesis (H1). Usually, the null hypothesis states that the two sample means are equal, and the alternative hypothesis states that the two sample means are not equal.

[0068] Null hypothesis (H0): μ1=μ2 (the two sample means are equal)

[0069] Alternative hypothesis (H1): μ1≠μ2 (the two sample means are not equal)

[0070] Collect data: Collect data for two samples and make sure the data meets the assumption of normal distribution. You can use a normal probability plot (Q-Qplot) or a statistical test (such as the Shapiro-Wilk test) to verify whether the data is approximately normally distributed;

[0071] Calculate t-statistic: Use the following formula to calculate t-statistic:

[0072] [t=\frac{\bar{X}_1-

[0073] \bar{X}_2}{\sqrt{\frac{S_1^2}{n_1}+\frac{S_2^2}{n_2}}}]

[0074] in:

[0075] (\bar{X}_1) and (\bar{X}_2) are the means of the two samples respectively.

[0076] (S_1) and (S_2) are the standard deviations of the two samples respectively.

[0077] (n_1) and (n_2) are the sizes of the two samples respectively.

[0078] To further optimize the solution, the process of building the Transformer model includes:

[0079] Construct an initial Transformer model with a multi-head attention mechanism, a first residual network, a feedforward neural network, and a second residual network connected in sequence;

[0080] Processing the output sequence of the Transformer model to obtain a processing model;

[0081] The fully connected layer is used as the output layer of the processing model to generate the Transformer model.

[0082] To further optimize the scheme, the expression of the multi-head attention mechanism is:

[0083] MHSA(x)=Concat(head1,…,head m )W O ;

[0084]

[0085] Among them, B rel is the relative position information matrix, which is obtained by multiplying the relative position distance by a specific factor; Q, K, and V are the query, index, and content vectors of the input data, respectively. are the weight matrices corresponding to the i-th head Q, K, and V, respectively, and W O is the weight matrix, m represents the number of heads; d K Represents the feature dimension.

[0086] To further optimize the scheme, the process of training the Transformer model based on the sample set includes:

[0087] Pre-training the Transformer model through the BERT pre-training model to obtain a pre-training model;

[0088] Acquire a sample set, perform iterative training on a pre-trained model based on the sample set, and terminate the iteration after the pre-trained model reaches automatic learning of input attribute data, operation data, and tension data, thereby obtaining the tension prediction model;

[0089] The loss function of the Transformer model is the cross entropy loss function.

[0090] Further optimizing the scheme, the process of iteratively training the pre-trained model based on the sample set further includes:

[0091] Dividing the sample set of the tension detection into a training set and a test set;

[0092] The pre-trained model is trained using the training set, a cross entropy loss function is selected, and the cross entropy loss function is minimized by a stochastic gradient descent algorithm to obtain a stochastic gradient descent model;

[0093] Using a validation set to validate the performance of the stochastic gradient descent model, to obtain a validation result;

[0094] Based on the verification result, the learning rate and the number of pooling layers of the stochastic gradient descent model are optimized to obtain a trained model.

[0095] Further optimizing the scheme, the data set to be calculated is input into the tension prediction model for calculation, and the process of obtaining the tension prediction curve includes:

[0096] Inputting the data set to be calculated into the tension prediction model for calculation to obtain a calculation result;

[0097] The calculation results are visually interpreted to obtain the tension prediction curve.

[0098] The preprocessed data set is input into the tension prediction model. The model will calculate the tension prediction value for a period of time in the future based on the learned historical data pattern. These predicted values ​​can be used to generate a tension prediction curve, which reflects the trend of tension changes during the prediction time.

[0099] Further optimizing the scheme, obtaining data points that do not conform to the normal operating range based on the tension prediction curve, and analyzing the source of tension anomaly based on the data points that do not conform to the normal operating range include:

[0100] Identifying outliers in the data distribution of the tension prediction curve based on the isolation forest;

[0101] Based on the generated tension prediction curve, data points that do not conform to the normal operating range can be identified. These points may indicate potential problems or anomalies that require further analysis. Judgment criteria may include tension values ​​exceeding predetermined upper and lower limits, or rapid changes in a short period of time.

[0102] Analyze the causal relationship between the abnormal point and the attribute data and the operation data to obtain an analysis result;

[0103] The analysis results are compiled into a report, which includes a detailed description of the anomaly, the presumed cause, and the suggested solution.

[0104] The actual observed tension data or the tension data predicted by the model is compared with the normal operating range. Any data point outside this range is marked as an anomaly. An isolation forest algorithm is used to identify outliers in the data distribution and the detected outliers are marked in the system.

[0105] For each marked abnormal data point, analyze the possible causes, including data backtracking: check the operation records, equipment status records, environmental conditions, etc. before and after the abnormality occurs. Correlation analysis: Analyze the correlation between the abnormal point and other variables to find possible cause and effect relationships. Organize the analysis results into a report, including a detailed description of the abnormality, the speculated cause, and the recommended solutions. According to the abnormal tension pattern, determine the possible type of fault. For example, if the tension suddenly increases, it may be that the chain is stuck or there is an increase in resistance caused by damaged components. In combination with the design and operation manual of the scraper conveyor, analyze the specific components or causes of the fault that may cause this tension change. Based on the analysis results, emergency measures may need to be taken to adjust the operating parameters or repair the equipment to prevent the abnormal situation from happening again.

[0106] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A real-time monitoring method for chain transmission tension of a scraper conveyor, characterized in that: The following steps are involved: Acquire the attribute data of the scraper conveyor, and acquire the operation data and tension data of the scraper conveyor in the working state and the non-working state through the sensor; Integrate the attribute data, operation data and tension data to obtain a data set to be calculated; Constructing a Transformer model, and training the Transformer model based on the sample set to obtain a tension prediction model; Inputting the data set to be calculated into the tension prediction model for calculation to obtain a tension prediction curve; Data points that do not conform to a normal operating range are acquired based on the tension prediction curve, and sources of tension anomalies are analyzed based on the data points that do not conform to a normal operating range.

2. The real-time monitoring method for chain transmission tension of a scraper conveyor according to claim 1 is characterized in that: The process of integrating the attribute data, operation data and tension data to obtain the data set to be calculated includes: Correcting erroneous or missing data points in the attribute data, operation data, and tension data to obtain a corrected data set; Synchronizing the timestamps in the corrected data set to generate aligned data; Merging the alignment data based on the device ID of the scraper conveyor to generate a merged data set; The merged data set is tested to obtain the data set to be calculated.

3. The real-time monitoring method for chain transmission tension of a scraper conveyor according to claim 2 is characterized in that: The process of testing the merged data set to obtain the data set to be calculated includes: Performing a normality test on the combined data set to obtain normally distributed data and non-normally distributed data; The normally distributed data are subjected to a t test, and the non-normally distributed data are subjected to a parameter rank sum test to obtain the data set to be calculated.

4. The method for real-time monitoring of chain transmission tension for a scraper conveyor according to claim 1, characterized in that: The process of building the Transformer model includes: Construct an initial Transformer model with a multi-head attention mechanism, a first residual network, a feedforward neural network, and a second residual network connected in sequence; Processing the output sequence of the Transformer model to obtain a processing model; The fully connected layer is used as the output layer of the processing model to generate the Transformer model.

5. The method for real-time monitoring of chain transmission tension for a scraper conveyor according to claim 4, characterized in that: The expression of the multi-head attention mechanism is: MHSA(x)=Concat(head1,…,head m )W O ; Among them, B rel is the relative position information matrix, which is obtained by multiplying the relative position distance by a specific factor; Q, K, and V are the query, index, and content vectors of the input data, respectively. are the weight matrices corresponding to the i-th head Q, K, and V, respectively, and W O is the weight matrix, m represents the number of heads; d K Represents the feature dimension.

6. The method for real-time monitoring of chain transmission tension for a scraper conveyor according to claim 1, characterized in that: The process of training the Transformer model based on the sample set includes: Pre-training the Transformer model through the BERT pre-training model to obtain a pre-training model; Acquire a sample set, perform iterative training on a pre-trained model based on the sample set, and terminate the iteration after the pre-trained model reaches automatic learning of input attribute data, operation data, and tension data, thereby obtaining the tension prediction model; The loss function of the Transformer model is the cross entropy loss function.

7. The method for real-time monitoring of chain transmission tension for a scraper conveyor according to claim 1, characterized in that: The process of iteratively training the pre-trained model based on the sample set also includes: Dividing the sample set of the tension detection into a training set and a test set; The pre-trained model is trained using the training set, a cross entropy loss function is selected, and the cross entropy loss function is minimized by a stochastic gradient descent algorithm to obtain a stochastic gradient descent model; Using a validation set to validate the performance of the stochastic gradient descent model, to obtain a validation result; Based on the verification result, the learning rate and the number of pooling layers of the stochastic gradient descent model are optimized to obtain a trained model.

8. The method for real-time monitoring of chain transmission tension for a scraper conveyor according to claim 1, characterized in that: The process of inputting the data set to be calculated into the tension prediction model for calculation to obtain the tension prediction curve includes: Inputting the data set to be calculated into the tension prediction model for calculation to obtain a calculation result; The calculation results are visually interpreted to obtain the tension prediction curve.

9. The method for real-time monitoring of chain transmission tension for a scraper conveyor according to claim 1, characterized in that: The process of obtaining data points that do not meet the normal operating range based on the tension prediction curve and analyzing the source of the tension anomaly based on the data points that do not meet the normal operating range includes: Identifying outliers in the data distribution of the tension prediction curve based on the isolation forest; Analyze the causal relationship between the abnormal point and the attribute data and the operation data to obtain an analysis result; The analysis results are compiled into a report, which includes a detailed description of the anomaly, the presumed cause, and the suggested solution.

Citation Information

Cited By

  • Scraper chain tension dynamic optimization method based on real-time load feedback

    CN121107023A