Fault prediction and health management system for belt conveyor

Abstract

The invention discloses a fault prediction and health management system for a belt conveyor, which comprises data acquisition and processing: a sensor arranged on the belt conveyor acquires corresponding original signals and inputs the original signals into a data acquisition module, the data acquisition module performs modulation and signal amplification processing on the original signals, and the original signals are transmitted to the data acquisition module; converting the signal into a digital signal, and generating diagnosis data; diagnosis data judgment: transmitting the diagnosis data to a judgment module for judgment, setting a data range according to the characteristics of the diagnosis data detected by each sensor before the diagnosis data is input, and when the diagnosis data is input, if the diagnosis data is in the preset data range, regarding the diagnosis data as a normal state, and otherwise, regarding the diagnosis data as a non-normal state; and if the diagnosis data is not in the preset data range, an abnormal state is considered, and the whole system can predict the fault of the belt conveyor, so that the influence of the fault on production is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of belt fault prediction systems, and in particular to a fault prediction and health management system for belt conveyors. Background Art

[0002] Belt conveyors are important equipment in industrial production and are widely used in mining, manufacturing and logistics industries. Their reliability and efficiency directly affect the operation of the production line. However, traditional maintenance methods often rely on regular inspections and manual testing, which makes it difficult to identify potential faults in a timely and accurate manner, resulting in unexpected downtime, production interruptions and high maintenance costs.

[0003] With the continuous advancement of industrial automation and intelligence, enterprises have put forward higher requirements for real-time monitoring and fault prediction of equipment. Advanced monitoring systems can obtain real-time operating data of conveyors through a variety of sensors, such as key parameters such as voltage, current, temperature and pressure. These data are processed by the data acquisition system, providing a reliable basis for fault diagnosis and prediction.

[0004] The introduction of intelligent fault prediction systems makes preventive maintenance possible. Such systems can issue early warnings before a fault occurs, greatly reducing equipment downtime and improving the continuity and efficiency of production lines. In addition, by optimizing resource allocation and maintenance strategies, companies can significantly reduce operating costs.

[0005] In this context, developing an efficient fault prediction and health management system is in line with the current development trend of industrial automation. Therefore, there is an urgent need for a system that can realize belt conveyor fault diagnosis to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide a fault prediction and health management system for a belt conveyor in order to better and more effectively solve the above problems.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] A fault prediction and health management system for a belt conveyor, comprising the following steps:

[0009] Step A, data acquisition and processing: The sensor installed on the belt conveyor will obtain the corresponding original signal and input it into the data acquisition module. The data acquisition module will modulate and amplify the original signal, convert it into a digital signal, and generate diagnostic data;

[0010] Step B, diagnosis data judgment: the diagnosis data is transmitted to the judgment module for judgment. Before the diagnosis data is input, the data range is set according to the characteristics of the diagnosis data detected by each sensor. When the diagnosis data is input, if the diagnosis data is within the preset data range, it is regarded as a normal state. If the diagnosis data is not within the preset data range, it is regarded as an abnormal state.

[0011] Step C, fault diagnosis: set up a fault information database, and train a fault diagnosis system based on the random forest strategy according to the fault information database, input the diagnosis data of the abnormal state into the fault diagnosis system for diagnosis, obtain the possibility of abnormality, and display the possible abnormality after the diagnosis is completed;

[0012] Step D, data update: manually determine whether the displayed abnormal alarm information and the possible abnormal conditions obtained through diagnosis are correct. If correct, label them. If incorrect, manually enter the correct label and store it in the fault information database.

[0013] Preferably, the specific steps of setting up the fault information database in step C are to collect fault information entries, each of which contains all monitored data types and their corresponding abnormality types.

[0014] Preferably, the specific steps of training the fault diagnosis system based on the random forest strategy according to the fault information database in step C are:

[0015] Step C1, extracting features: extracting B features from each fault information entry, each feature having a corresponding weight parameter;

[0016] Step C2, training decision trees: select a sub-dataset of size M from the fault information database with replacement, repeat K times to obtain K sub-datasets, use the K sub-datasets as training sets, and train to obtain K decision trees, where the K decision trees are estimators of the fault diagnosis system.

[0017] Preferably, in step C, the specific steps of inputting the abnormal state diagnosis data into the fault diagnosis system for diagnosis are:

[0018] Step C3, integrated calculation: Input the diagnostic data into the fault diagnosis system, submit the diagnostic data to each decision tree for prediction, and integrate the decision results of all decision trees. i , the specific calculation formula is:

[0019]

[0020] Where j = 1, 2, ... B, i refers to the i-th possibility, f j (x) is the calculation result of a single decision tree for the possible occurrence probability of the current abnormality.

[0021] Preferably, in step D, a label is attached, where the label is a combination of a single fault information or multiple fault information.

[0022] Preferably, in step C2, the decision tree does not set weights.

[0023] Preferably, in step C2, b features are used for training during the training process of each decision tree, where b < B, and the b features are randomly selected from the overall B features.

[0024] The beneficial effects of the present invention are:

[0025] 1. Real-time monitoring and data processing:

[0026] The mechanical and electrical full-process state data of the belt conveyor are collected in real time through various types of sensors (such as voltage, current, thermocouple, pressure sensors, etc.);

[0027] The data acquisition module modulates, amplifies, and converts the original signal into a digital signal to ensure the accuracy and consistency of the data.

[0028] 2. Efficient data analysis and storage:

[0029] A wired communication connection is adopted to ensure the reliability of data transmission;

[0030] A local database is established to store sensor data, which is convenient for subsequent analysis and fault judgment.

[0031] 3. Intelligent fault diagnosis:

[0032] A fault diagnosis system based on the random forest strategy is used to intelligently diagnose abnormal data, improving the diagnosis speed and accuracy;

[0033] The continuous update of the fault information database improves the self-learning ability of the system.

[0034] 4. Flexible data range setting:

[0035] It is allowed to adjust the normal range of diagnostic data according to the actual operation situation, improving the adaptability and accuracy of the system.

[0036] 5. Continuous optimization and model update:

[0037] Through manual confirmation and label update, the fault diagnosis model is continuously optimized to enhance the prediction ability;

[0038] The diversity and integration of the random forest model are used to improve the stability of the diagnostic system and avoid overfitting.

[0039] 6. Comprehensive fault identification ability:

[0040] It can identify single faults and compound faults, and improve the system's fault handling capabilities. DETAILED DESCRIPTION

[0041] The present invention will be further described below in conjunction with the specification.

[0042] A fault prediction and health management system for a belt conveyor of the present invention comprises the following steps:

[0043] Step A, data acquisition and processing: The sensor installed on the belt conveyor will obtain the corresponding original signal and input it into the data acquisition module. The data acquisition module will modulate and amplify the original signal, convert it into a digital signal, and generate diagnostic data;

[0044] The sensors include but are not limited to voltage sensors, current sensors, thermocouples, pressure sensors, etc. Each sensor is responsible for collecting one or more signals. The locations where the sensors are set include drum motors, reducers, roller devices, etc. Through these sensors, the system can receive the status data of the entire mechanical and electrical process, and store and analyze these data.

[0045] The sensor and the data acquisition module are connected by wired communication. After that, the data acquisition module will transmit the converted digital signal to the central monitoring system for further processing. A local database for storing sensor data is established, and the converted data is stored in the database for subsequent judgment and analysis.

[0046] Step B, diagnosis data judgment: the diagnosis data is transmitted to the judgment module for judgment. Before the diagnosis data is input, the data range is set according to the characteristics of the diagnosis data detected by each sensor. When the diagnosis data is input, if the diagnosis data is within the preset data range, it is regarded as a normal state. If the diagnosis data is not within the preset data range, it is regarded as an abnormal state.

[0047] The data range can be changed according to the actual operating conditions.

[0048] Step C, fault diagnosis: set up a fault information database, and train a fault diagnosis system based on the random forest strategy according to the fault information database, input the diagnosis data of the abnormal state into the fault diagnosis system for diagnosis, obtain the possibility of abnormality, and display the possible abnormality after the diagnosis is completed;

[0049] Step D, data update: manually determine whether the displayed abnormal alarm information and the possible abnormal conditions obtained through diagnosis are correct. If correct, label them. If incorrect, manually enter the correct label and store it in the fault information database.

[0050] Among them, if the correct label is not manually entered and stored in the fault information database, the data will be permanently saved, and the data will update the random forest model to improve its prediction and diagnosis capabilities.

[0051] Specifically, the specific steps for setting up the fault information database in step C are to collect fault information entries, each of which contains all monitored data types and their corresponding abnormality types. The collection process includes but is not limited to data collected in advance based on actual conditions, and data after an abnormality occurs during the operation of the belt conveyor belt.

[0052] Specifically, in step C, the specific steps of training the fault diagnosis system based on the random forest strategy according to the fault information database are as follows:

[0053] Step C1, extracting features: extracting B features from each fault information entry, each feature having a corresponding weight parameter;

[0054] Step C2, training decision trees: select a sub-dataset of size M from the fault information database with replacement, repeat K times to obtain K sub-datasets, use the K sub-datasets as training sets, and train to obtain K decision trees, where the K decision trees are estimators of the fault diagnosis system.

[0055] Specifically, in step C, the diagnostic data of the abnormal state is input into the fault diagnosis system for diagnosis. The specific steps are:

[0056] Step C3, integrated calculation: Input the diagnostic data into the fault diagnosis system, submit the diagnostic data to each decision tree for prediction, and integrate the decision results of all decision trees. i , the specific calculation formula is:

[0057]

[0058] Where j = 1, 2, ... B, i refers to the i-th possibility, f j (x) is the calculation result of a single decision tree for the possible occurrence probability of the current abnormality.

[0059] Specifically, a label is added in step D, wherein the label is a single fault information or a combination of multiple fault information, so that the overall ability to diagnose complex faults is improved while maintaining the single abnormality diagnosis capability.

[0060] Specifically, the decision tree in step C2 does not set weights, which can simplify the entire calculation process. Moreover, the random forest creates diverse decision trees by randomly selecting feature and data subsets. This diversity itself is a guarantee of stability. The strength of the random forest lies in integrating multiple decision trees to improve the overall accuracy. There is no need to set weights for individual trees, and not setting weights helps reduce the risk of overfitting because each tree contributes equally to the overall prediction.

[0061] Specifically, during the training process of each decision tree in step C2, b features are used for training, where b < B, and the b features are randomly selected from the overall B features.

[0062] To better understand this solution, three specific embodiments are presented in combination with the actual situation.

[0063] Embodiment 1

[0064] Combined fault detection (motor overheating and belt looseness)

[0065] Data acquisition:

[0066] The thermocouple sensor monitors the motor temperature, and the pressure sensor monitors the belt tension.

[0067] Data processing:

[0068] The data acquisition module modulates, amplifies the temperature and tension signals, converts them into digital signals, and stores them in the local database.

[0069] Diagnostic judgment:

[0070] The system compares the temperature and tension data with their respective set ranges, and if exceeded, marks them as abnormal.

[0071] Fault diagnosis:

[0072] The random forest model comprehensively analyzes the abnormal data to identify and confirm whether there are both motor overheating and belt looseness at the same time.

[0073] Data update:

[0074] The operator verifies the combined fault alarm. If it is correct, a combined label is marked. Otherwise, the correct label is input and the fault database is updated.

[0075] Embodiment 2

[0076] Belt looseness detection

[0077] Data acquisition:

[0078] The pressure sensor is installed at the idler device to monitor the belt tension.

[0079] Data processing:

[0080] The signal is modulated and amplified, converted into a digital signal for storage.

[0081] Diagnosis:

[0082] If the tension is lower than the normal range, it is judged as an abnormal state.

[0083] Troubleshooting:

[0084] Using the fault information database, the random forest system predicts the possibility of belt loosening.

[0085] Data update:

[0086] Manually check and confirm the diagnosis results, and update the database to improve the model accuracy.

[0087] Example 3

[0088] Abnormal conveyor current

[0089] Data collection:

[0090] The current sensor is installed on the drum motor to collect current data in real time.

[0091] Data processing:

[0092] The signal is converted into digital form and input into the central monitoring system.

[0093] Diagnosis:

[0094] If the current is higher or lower than the set range, it is marked as abnormal.

[0095] Troubleshooting:

[0096] The random forest model analyzes current anomalies and infers possible mechanical failures.

[0097] Data update:

[0098] Continuously optimize fault diagnosis capabilities through manual confirmation and label correction.

[0099] In summary, the fault prediction and health management system for a belt conveyor of the present invention has the following effects:

[0100] 1. Real-time monitoring and data processing:

[0101] The mechanical and electrical status data of the belt conveyor are collected in real time through various types of sensors (such as voltage, current, thermocouple, pressure sensor, etc.);

[0102] The data acquisition module modulates, amplifies and converts the original signal into a digital signal to ensure the accuracy and consistency of the data.

[0103] 2. Efficient data analysis and storage:

[0104] Adopt wired communication connection to ensure the reliability of data transmission;

[0105] Establish a local database to store sensor data for subsequent analysis and fault diagnosis.

[0106] 3. Intelligent fault diagnosis:

[0107] Use a fault diagnosis system based on random forest strategy to intelligently diagnose abnormal data and improve diagnosis speed and accuracy;

[0108] The continuous updating of the fault information database improves the self-learning ability of the system.

[0109] 4. Flexible data range setting:

[0110] It allows the normal range of diagnostic data to be adjusted according to actual operating conditions, improving the adaptability and accuracy of the system.

[0111] 5. Continuous optimization and model update:

[0112] Through manual confirmation and label update, the fault diagnosis model is continuously optimized to enhance the prediction capability;

[0113] The diversity and integration of random forest models are used to improve the stability of the diagnostic system and avoid overfitting.

[0114] 6. Comprehensive fault identification capability:

[0115] It can identify single faults and compound faults, and improve the system's fault handling capabilities.

[0116] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A fault prediction and health management system for a belt conveyor, characterized in that: including the following steps, Step (A), data acquisition and processing: The sensors set on the belt conveyor will obtain the corresponding original signals and input them into the data acquisition module. The data acquisition module will perform modulation and demodulation and signal amplification processing on the original signals, convert them into digital signals, and generate diagnostic data; Step (B), judgment of diagnostic data: The diagnostic data is transmitted to the judgment module for judgment. Before inputting the diagnostic data, the data range is set according to the characteristics of the diagnostic data detected by each sensor. When the diagnostic data is input, if the diagnostic data is within the preset data range, it is regarded as the normal state; if the diagnostic data is not within the preset data range, it is regarded as the abnormal state; Step (C), fault diagnosis: A fault information database is set up, and a fault diagnosis system based on the random forest strategy is trained according to the fault information database. The diagnostic data in the abnormal state is input into the fault diagnosis system for diagnosis to obtain the possible abnormality. After the diagnosis is completed, the possible abnormal situation is displayed; Step (D), data update: Manually judge whether the displayed abnormal alarm information and the possible abnormal situation obtained through diagnosis are correct. If it is correct, label it. If it is incorrect, manually input the correct label and store it in the fault information database.

2. A fault prediction and health management system for a belt conveyor according to claim 1, characterized in that: The specific steps for setting up the fault information database in Step (C) are as follows: Collect fault information entries. Each fault information entry contains all the monitored data types and their corresponding abnormal types.

3. A fault prediction and health management system for a belt conveyor according to claim 2, characterized in that: The specific steps for training the fault diagnosis system based on the random forest strategy according to the fault information database in Step (C) are as follows: Step (C1), feature extraction: Extract B features from each fault information entry, and each feature has a corresponding weight parameter; Step (C2), training decision trees: Select a sub-dataset with a size of M from the fault information database with replacement, repeat K times to obtain K sub-datasets. Use the K sub-datasets as the training set to train K decision trees. The K decision trees are the estimators of the fault diagnosis system.

4. A fault prediction and health management system for a belt conveyor according to claim 2, characterized in that: The specific steps for inputting the diagnostic data in the abnormal state into the fault diagnosis system for diagnosis in Step (C) are as follows: Step (C3), integrated calculation: input the diagnostic data into the fault diagnosis system, submit the diagnostic data to each decision tree for prediction, and integrate the decision results of all decision trees for each possible abnormality y i , the specific calculation formula is: Where j = 1, 2, ... B, i refers to the i-th possibility, f j (x) is the calculation result of a single decision tree for the possible occurrence probability of the current abnormality.

5. The fault prediction and health management system for a belt conveyor according to claim 1, characterized in that: In Step (D), label it, where the label is a combination of a single fault information or multiple fault information.

6. A fault prediction and health management system for a belt conveyor according to claim 3, characterized in that: In Step (C2), no weights are set for the decision trees.

7. The fault prediction and health management system for a belt conveyor according to claim 3, characterized in that: In Step (C2), b features are used for training during the training process of each decision tree, where b < B, and the b features are randomly selected from the overall B features.

Images