PHM model system of cement production management and control platform

Abstract

The invention provides a cement production management and control platform PHM model system, and the system comprises the steps: obtaining the monitoring data and production management data of a cement production process, and enabling the monitoring data and production management data to form an original data set; performing data cleaning and preprocessing on the original data set, and removing noise and abnormal values to obtain a standardized data set; inputting the feature component into a preset deep learning model, and outputting a prediction result of the equipment state; the prediction result is compared with a set threshold range, and if the prediction result exceeds the threshold range, early warning information is generated; classifying the early warning information in combination with historical fault data and current production environment parameters, and distinguishing normal production fluctuation and potential faults; if the classification result is a potential fault, a fault mode is analyzed, a fault type is determined, a pre-training model is finely adjusted by using a fault sample, and the generalization ability of the model is improved; new data are continuously collected in the production process, the model is periodically updated, and the accuracy of model fault early warning is improved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a PHM model system for a cement production management and control platform. Background Art

[0002] Fault early warning for cement production equipment is a complex technical challenge. A large amount of real-time data is generated during the operation of the equipment. These data contain important information about the equipment status, but how to extract effective features from them and accurately predict potential faults becomes a key issue. Traditional methods often rely on fixed thresholds and simple rules, which are difficult to cope with complex and changing production environments. Although advanced machine learning algorithms have powerful pattern recognition capabilities, they face problems such as unstable data quality and poor model generalization in practical applications. In addition, different types of faults may show similar symptoms, and how to accurately distinguish between normal fluctuations and potential faults is also a major problem. Even if potential faults are successfully identified, how to issue warnings in a timely and effective manner and take corresponding measures to minimize production losses is also an urgent problem to be solved. These challenges are interrelated and together constitute the core technical difficulties of fault early warning for cement production equipment. How to build an accurate, reliable, and adaptable fault early warning system using limited historical data and real-time monitoring information in a complex and changing production environment is a major challenge facing this field. Summary of the invention

[0003] The present invention provides a PHM model system for a cement production control platform, which mainly includes: Acquire monitoring data and production management data of the cement production process, wherein the monitoring data and production management data constitute an original data set; perform data cleaning and preprocessing on the original data set to remove noise and outliers and obtain a standardized data set; perform feature extraction on the standardized data set to obtain feature indicators, and if the feature indicators show a nonlinear change trend, perform feature decomposition to obtain feature components; input the feature components into a preset deep learning model to output a prediction result of the equipment status; compare the prediction result with a set threshold range, and if the threshold range is exceeded, generate an early warning message; classify the early warning message in combination with historical fault data and current production environment parameters to distinguish between normal production fluctuations and potential faults; if the classification result is a potential fault, analyze the fault mode, determine the fault type, and use the fault sample to fine-tune the pre-trained model to improve the generalization ability of the model; continuously collect new data during the production process, update the model regularly, and improve the accuracy of the model fault warning.

[0004] Furthermore, the acquisition of monitoring data and production management data of the cement production process includes: collecting equipment operation data through vibration sensors and temperature sensors deployed on cement production equipment, combining historical fault records to obtain a training data set with equipment status labels, and the status labels include normal operation and potential faults.

[0005] Furthermore, the preset deep learning model includes: training a deep learning model that integrates a convolutional neural network and a long short-term memory network based on a training data set with equipment status labels, using the convolutional neural network to extract the spatial features of the equipment operation data, and using the long short-term memory network to capture the long-term dependencies of the time series data, setting the prediction time window to 1 hour, and obtaining a deep learning model capable of predicting the status of the equipment within the next 1 hour.

[0006] Furthermore, the prediction result is compared with the set threshold range, including: setting a root mean square error threshold upper limit parameter and a root mean square error threshold lower limit parameter to form a root mean square error threshold range; obtaining the probability of bearing wear occurring in the device outputting the prediction result, if the predicted bearing wear probability is greater than the root mean square error threshold upper limit parameter or less than the root mean square error threshold lower limit parameter, then the comparison result is output as exceeding the root mean square error threshold range.

[0007] Furthermore, the generation of early warning information includes: if the comparison result output exceeds the root mean square error threshold range, the alarm trigger condition is met, and according to the alarm trigger condition, an alarm information content including the equipment number, predicted time, fault type as bearing wear and predicted probability is generated, and a multi-level alarm mechanism is started to determine the alarm level according to the predicted bearing wear probability.

[0008] Furthermore, the initiation of the multi-level alarm mechanism includes: judging whether the equipment has a potential risk of bearing wear failure based on the alarm information content, and using a preset personnel notification method to determine the mobile phone numbers and email addresses of the maintenance team leader and the workshop director; sending the alarm information content to the maintenance team leader via SMS notification to obtain a notification of the potential risk of bearing wear failure of the equipment; at the same time, if the alarm level is level one, sending the alarm information content to the workshop director via email notification to obtain a notification of the potential risk of bearing wear failure of the equipment.

[0009] Furthermore, the maintenance team leader determines the specific time point and equipment component corresponding to the abnormal data segment based on the received SMS notification and the abnormal data segment in the prediction result highlighted by the self-attention mechanism, and obtains an equipment maintenance plan including troubleshooting steps and required spare parts information.

[0010] Furthermore, if the workshop director receives more than three alarms of the same type in one day based on the email notification received, the alarm merging mechanism is triggered to generate a merged alarm containing the equipment numbers and time information of all triggered alarms, which is sent to the production manager via email notification to obtain notification of the potential risk of large-scale equipment failure.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a method for early warning of cement production equipment failure. The method collects, cleans and preprocesses real-time monitoring data, combines wavelet transform to perform feature decomposition, and inputs the processed data into a deep learning model to predict the equipment status. When the prediction result exceeds a preset threshold, the support vector machine algorithm is used to distinguish normal fluctuations from potential failures in combination with historical failure data and current environmental parameters. For potential failures, the present invention uses transfer learning technology to fine-tune the pre-trained model to improve generalization capabilities. The accuracy of the early warning is ensured by continuously collecting new data and regularly updating the model. When a potential failure risk is detected, a multi-level alarm mechanism is activated to promptly notify relevant personnel. The present invention realizes real-time monitoring, fault early warning and intelligent diagnosis of the status of cement production equipment, thereby improving production efficiency and equipment reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention is a flow chart of a PHM model system for a cement production control platform.

[0013] Figure 2 It is a schematic diagram of a PHM model system of a cement production control platform of the present invention.

[0014] Figure 3 This is another schematic diagram of a PHM model system for a cement production control platform of the present invention.

[0015] Figure 4 This is another schematic diagram of a PHM model system for a cement production control platform of the present invention.

[0016] Figure 5 This is another schematic diagram of a PHM model system for a cement production control platform of the present invention.

[0017] Figure 6 This is another schematic diagram of a PHM model system for a cement production control platform of the present invention.

[0018] Figure 7 This is another schematic diagram of a PHM model system for a cement production control platform of the present invention.

[0019] Figure 8 This is another schematic diagram of a PHM model system for a cement production control platform of the present invention.

[0020] Fig. 9 This is another schematic diagram of a PHM model system for a cement production control platform of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] like Figure 1-9 In this embodiment, a PHM model system of a cement production control platform may specifically include: S101. Acquire data from real-time monitoring and production management modules in the cement production process, remove noise and outliers through data cleaning and preprocessing, and obtain a standardized data set.

[0023] S102: If the characteristic index in the standardized data set shows a nonlinear change trend, a wavelet transform method is used to perform further characteristic decomposition to obtain more refined characteristic components.

[0024] S103: Input the characteristic components into a preset deep learning model integrating a convolutional neural network and a long short-term memory network, and output a prediction result of the device state.

[0025] S104: According to the prediction result, a root mean square error threshold range is set to determine whether the device is in a potential failure state.

[0026] S105. If the prediction result exceeds the root mean square error threshold range, a warning message is generated. The warning message is combined with historical fault data and current production environment parameters, and is classified using a support vector machine algorithm to distinguish normal production fluctuations from potential faults.

[0027] S106. If the classification result is a potential fault, the fault mode is analyzed to determine the fault type, and transfer learning technology is used to fine-tune the pre-trained model using limited fault samples to improve the generalization ability of the model.

[0028] S107. Continuously collect new data during the actual production process, and regularly update and optimize the model online to ensure the model's sensitivity to changes in equipment status and the accuracy of fault warnings.

[0029] S108. Based on the early warning information, a multi-level alarm mechanism is activated, and the alarm information is sent to the maintenance team leader, workshop director, production manager and other relevant personnel via SMS and email to obtain notification of potential equipment failure risks.

[0030] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A PHM model system for cement production control platform, characterized in that: include: Acquire monitoring data and production management data of the cement production process, wherein the monitoring data and production management data constitute an original data set; Performing data cleaning and preprocessing on the original data set to remove noise and outliers to obtain a standardized data set; Performing feature extraction on the standardized data set to obtain feature indicators, and if the feature indicators show a nonlinear change trend, performing feature decomposition to obtain feature components; Inputting the feature components into a preset deep learning model and outputting a prediction result of the device state; Compare the prediction results with the set threshold range, and generate warning information if it exceeds the threshold range; Combine historical fault data and current production environment parameters to classify the warning information and distinguish between normal production fluctuations and potential faults; If the classification result is a potential fault, the fault mode is analyzed to determine the fault type, and the pre-trained model is fine-tuned using the fault samples to improve the model's generalization ability; Continuously collect new data during the production process, update the model regularly, and improve the accuracy of model fault warnings.

2. The PHM model system of cement production control platform according to claim 1, characterized in that: The acquisition of monitoring data and production management data of the cement production process includes: Vibration sensors and temperature sensors deployed on cement production equipment are used to collect equipment operation data. Combined with historical fault records, a training dataset with equipment status labels is obtained. The status labels include normal operation and potential faults.

3. The PHM model system of cement production control platform according to claim 1, characterized in that: The preset deep learning model includes: Based on the training data set with equipment status labels, a deep learning model that integrates convolutional neural networks and long short-term memory networks is trained. The convolutional neural network is used to extract the spatial features of the equipment operation data, and the long short-term memory network is used to capture the long-term dependencies of time series data. The prediction time window is set to 1 hour, and a deep learning model with the ability to predict the equipment status within the next 1 hour is obtained.

4. The PHM model system of cement production control platform according to claim 1, characterized in that: The step of comparing the prediction result with a set threshold range includes: Setting a root mean square error threshold upper limit parameter and a root mean square error threshold lower limit parameter to form a root mean square error threshold range; The probability of bearing wear of the device whose prediction result is output is obtained. If the predicted bearing wear probability is greater than the upper limit parameter of the root mean square error threshold or less than the lower limit parameter of the root mean square error threshold, the comparison result is output as exceeding the root mean square error threshold range.

5. The cement production control platform PHM model system according to claim 1, characterized in that: The generating of early warning information includes: If the comparison result output is beyond the root mean square error threshold range, the alarm trigger condition is met. According to the alarm trigger condition, an alarm information content including the equipment number, predicted time, fault type as bearing wear and predicted probability is generated, and a multi-level alarm mechanism is started to determine the alarm level according to the predicted bearing wear probability.

6. The cement production control platform PHM model system according to claim 5, characterized in that: The multi-level alarm mechanism is initiated, including: Based on the alarm information, determine if the equipment has a potential risk of bearing wear failure, use the preset personnel notification method, and determine the mobile phone number and email address of the maintenance team leader and workshop director; The alarm information is sent to the maintenance team leader through SMS notification, and the team leader is notified of the potential risk of bearing wear failure of the equipment. At the same time, if the alarm level is level one, the alarm information is sent to the workshop director through email notification, and the workshop director is notified of the potential risk of bearing wear failure of the equipment.

7. The PHM model system of cement production control platform according to claim 6, characterized in that: The maintenance team leader determines the specific time point and equipment component corresponding to the abnormal data segment based on the received SMS notification and the abnormal data segment in the prediction result highlighted by the self-attention mechanism, and obtains an equipment maintenance plan including troubleshooting steps and required spare parts information.

8. The PHM model system of cement production control platform according to claim 6, characterized in that: If the workshop director receives more than three alarms of the same type in one day based on the email notification received, the alarm merging mechanism will be triggered to generate a merged alarm containing the equipment numbers and time information of all the equipment that triggered the alarms, and send it to the production manager via email notification to obtain notification of the potential risk of large-scale equipment failure.

Images