A predictive maintenance method and maintenance system for controllers

Abstract

This invention discloses a predictive maintenance method and system for controllers, belonging to the field of electromechanical maintenance technology. The method includes: S1, collecting native controller data and ambient temperature; S2, preprocessing the collected data, including outlier removal, missing value imputation, extraction of degradation-sensitive features, and normalization; S3, constructing a controller health assessment model using a weighted multi-index fusion algorithm to calculate the controller's current health; S4, classifying controller health levels based on the comprehensive health score and outputting multi-level warnings and corresponding maintenance suggestions; and S5, sending the warning information and corresponding maintenance suggestions to a display for visual reminders.

Description

Technical Field

[0001] This invention relates to the field of mechanical and electronic maintenance technology, specifically to a predictive maintenance method and system for controllers. Background Technology

[0002] Construction machinery (excavators, loaders, cranes, etc.) operate under harsh conditions of strong vibration, wide voltage fluctuation, high load, and large temperature difference for a long time. The controller is constantly subjected to electrical stress, thermal stress, and vibration stress, which can easily lead to gradual and insidious degradation such as power circuit aging, power device deterioration, CAN communication attenuation, CPU computing power decline, and output accuracy drift. It often stops suddenly without obvious warning, causing serious economic losses.

[0003] Chinese patent application CN119128436A discloses a method, system, and related equipment for predicting faults in engineering machinery. It proposes a multi-dimensional data fusion prediction model, whose core innovation lies in incorporating not only traditional vehicle operation data but also transportation data, thus providing a more comprehensive picture of equipment operating conditions and solving the problem of low prediction accuracy caused by a single data source. In terms of model construction, the patent employs a complex Stacking ensemble strategy, using Naive Bayes, CatBoost, SVM, and Random Forest models as base models and logistic regression as the meta-model to improve prediction accuracy and robustness. Furthermore, after removing outliers and filling in missing values ​​in historical indicator data, the method first classifies samples using fault data annotation, then filters features highly correlated with faults through correlation analysis, and finally uses oversampling and undersampling methods to address the imbalance between minority and majority class samples, thereby constructing a model capable of effectively outputting fault prediction results.

[0004] Chinese patent application CN121563471A discloses a health management method and device for construction machinery equipment. This method and device addresses the problems of over-maintenance and maintenance delays inherent in the traditional "periodic maintenance" model. The solution first retrieves monitoring indicators associated with various components of the construction machinery equipment from a pre-built indicator library. Then, through industrial IoT sensors and a data acquisition system, it acquires real-time monitoring data of these indicators as the data basis for health status assessment. Once the data is ready, the system accesses a pre-built maintenance knowledge base, which stores standardized maintenance measures corresponding to different components and monitoring indicators. Based on real-time monitoring data and system-preset risk level thresholds, the system automatically calculates the current health risk level of each component and determines whether this risk level falls within the "maintenance risk level range." If the triggering condition is met, the system retrieves the corresponding maintenance measures from the maintenance knowledge base and pushes them to the maintenance engineer's mobile terminal or the equipment management backend. This solution realizes a shift from time-cycle-based passive maintenance to real-time status-based predictive proactive maintenance.

[0005] The drawback is:

[0006] 1. Chinese patent application CN119128436A primarily predicts faults in engines and hydraulic systems. However, the electronic controller, the core of the system's decision-making, suffers from latent degradation issues such as power supply aging and communication attenuation, which remain unmonitored. A sudden controller failure could lead to system-wide paralysis. Furthermore, the Stacking integration model employed in this patent incurs significant memory overhead, while construction machinery typically relies on resource-constrained MCUs. The inference response capability of this solution in an airborne environment remains to be verified.

[0007] 2. Chinese patent application CN121563471A identifies the maintenance knowledge base and risk thresholds as the core of the system, the quality of which directly determines the effectiveness of the system's output. Building a knowledge base and risk thresholds covering construction machinery parts and complex working conditions requires extensive industry experience and a large amount of historical data, which is highly limiting and cannot be quickly implemented in entirely new scenarios without historical data. Furthermore, this solution is highly dependent on the quality of sensor data and the stability of the acquisition system, while the harsh working environment of construction machinery poses a severe challenge to the continuity and reliability of data acquisition.

[0008] 3. It is highly dependent on cloud computing, and cannot make real-time predictions in scenarios with no network or poor signal, such as mines and construction sites. It also has the risks of large data transmission delays, high traffic costs, and information security risks. Summary of the Invention

[0009] The present invention addresses the technical problems mentioned in the background section above. The first aspect provides a predictive maintenance method for controllers, optimizing data denoising, normalization, and feature extraction algorithms to improve the accuracy of health assessment and early warning. Simultaneously, it constructs a multi-dimensional weighted health assessment model, designs dynamic thresholds and adaptive weights based on the harsh working conditions of engineering machinery, and outputs tiered maintenance recommendations, thereby achieving a shift from passive maintenance to proactive predictive maintenance and meeting the needs of reliable operation and maintenance throughout the controller's entire lifecycle.

[0010] The second aspect provides a maintenance system that fully reuses core data collected by the controller itself, such as voltage, current, chip temperature, CPU load, CAN bus status, and frequency of occasional faults. Combined with ambient temperature, it achieves zero-cost data collection and multi-dimensional coverage. Relying on edge computing architecture, it performs local real-time calculations to predict health status.

[0011] According to the above-mentioned objectives of the present invention, the first aspect of the present invention provides a predictive control method for a controller, comprising the following steps:

[0012] Includes the following steps:

[0013] S1. Collect native data from the controller and ambient temperature;

[0014] S2. Preprocess the collected data from each source, including outlier removal, missing value imputation, extraction of degradation-sensitive features, and normalization.

[0015] S3. A weighted multi-index fusion algorithm is used to construct a controller health assessment model and calculate the current health of the controller.

[0016] S4. Based on the comprehensive health score, classify the controller health level and output multi-level early warnings and corresponding maintenance suggestions.

[0017] Furthermore, in step S1, the controller's native data includes the internal power supply voltage. Chip temperature CPU utilization Memory usage Communication error frame rate Frequency of historical failures runtime The sampling frequency is 1 time / minute; ambient temperature .

[0018] Furthermore, in step S2, the specific methods of data preprocessing include:

[0019] S21. Calculate the mean for each data stream. and standard deviation ,use In principle, if the data satisfy If the value is missing, it is identified as an outlier and removed. Linear interpolation is then used to fill in the missing or removed data.

[0020] S22. Extract degradation-sensitive features from the cleaned data, including calculating voltage fluctuations. Chip temperature rise Operating aging factors Failure rate ;in, , , , ;

[0021] S23. Using the min-max normalization method, the data of each dimension are mapped to the [0,1] interval. The calculation formula is as follows: .

[0022] Furthermore, in step S3, the specific steps for constructing the controller health assessment model include:

[0023] S31. Based on the statistical data of engineering machinery failures, the failure mechanism of the controller and expert experience, determine the weight of each evaluation index, and define the health threshold and failure threshold.

[0024] S32. Calculate the single-indicator health score and map the normalized data of each indicator to a single-indicator health score of 0 to 100.

[0025] S33. Weighted and integrated health scores of all individual indicators are used to obtain the comprehensive health score H.

[0026] Furthermore, in step S32, the formula for calculating the health score of each individual indicator is as follows:

[0027] Voltage stability health score ;

[0028] in, The voltage stability data is after preprocessing. The healthy threshold for voltage stability, This is the failure threshold for voltage stability;

[0029] Chip temperature rise health score ;

[0030] in, This is the preprocessed chip temperature rise data. This refers to the chip's temperature rise health threshold. This is the chip temperature rise failure threshold;

[0031] Failure rate and health score ;

[0032] in, The failure rate, The failure threshold is the failure rate, and the health threshold is 0.

[0033] CPU utilization health score ;

[0034] in, For CPU utilization, The healthy threshold for CPU utilization. The failure threshold for CPU utilization;

[0035] Memory usage health score ;

[0036] in, For memory usage, This represents a healthy threshold for memory usage. This is the failure threshold for memory usage.

[0037] Communication error rate health score ;

[0038] in, For communication error rate, The failure threshold for communication error rate;

[0039] Running aging factor health points ;

[0040] in, The value of the aging factor after pretreatment.

[0041] Furthermore, in step S33, the formula for calculating the overall health score is as follows:

[0042] ,in, As a weight for voltage stability, As a weight for chip temperature rise, To run the weights of the aging factor, As the weight of the failure rate, As a weight for CPU utilization, As a weight for memory usage, The weights are the communication error rate weights, and the sum of the weights is 1.

[0043] Furthermore, in step S4, the health level classification and corresponding warning levels are as follows:

[0044] Health level 80-100: The controller is operating normally and there are no warnings.

[0045] Sub-health level 60-79: Level 1 warning, preventive maintenance recommended;

[0046] Fault warning level 40-59: Level 2 alarm, key maintenance recommended;

[0047] Failure level 0-39: Level 3 emergency alarm, it is recommended to immediately stop the machine for inspection and replacement of the controller.

[0048] Furthermore, it also includes: S5, sending warning information and corresponding maintenance suggestions to the display to achieve visual reminders.

[0049] A second aspect of the present invention discloses a maintenance system, including a controller and a display communicatively connected thereto;

[0050] The controller includes a data acquisition module, a data preprocessing module, a health assessment module, and an early warning decision module;

[0051] The monitor is used to display data and provide visual reminders.

[0052] Furthermore, the data acquisition module is used to collect native data from the controller and ambient temperature in real time.

[0053] Furthermore, the data preprocessing module cleans each data stream, removes outliers, fills in missing values, extracts degradation-sensitive features from the cleaned data, and uses min-max normalization to normalize these degradation-sensitive feature data.

[0054] Furthermore, the health assessment module uses a weighted multi-indicator fusion algorithm to construct a controller health assessment model and calculate the controller's current health score.

[0055] Furthermore, the early warning decision module classifies the controller's health level based on the comprehensive health score and outputs multi-level early warnings and corresponding maintenance suggestions.

[0056] Furthermore, the early warning decision module encapsulates the information requiring early warning according to the predefined CAN protocol and sends it to the display via the CAN bus; the display automatically retrieves the corresponding prompt interface according to the early warning level and issues visual reminders to the user in the form of pop-ups and maintenance suggestion text.

[0057] The advantages of this invention compared to the prior art are:

[0058] 1. By utilizing the controller's own data and ambient temperature at the edge, the health status of the controller can be assessed locally in real time. Without relying on cloud computing, early warning decisions can be made in milliseconds, avoiding the impact of network interruptions and delays, and significantly improving the reliability and safety of vehicle operation.

[0059] 2. By monitoring the thermal stress and aging trend of the controller, potential faults can be identified in advance and early warning information and specific maintenance suggestions can be pushed to the display, realizing the transformation from post-maintenance to pre-maintenance, effectively avoiding equipment downtime, project interruption or even safety accidents caused by sudden controller failure, and effectively reducing project time and economic losses caused by unplanned downtime.

[0060] 3. Dynamically develop maintenance plans based on the actual health status of the controller to reduce unnecessary manual inspections, improve resource utilization efficiency, and effectively reduce maintenance costs and resource waste.

[0061] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of the controller predictive maintenance system framework according to an embodiment of the present invention;

[0063] Figure 2 This is a schematic diagram of the maintenance method according to an embodiment of the present invention. Detailed Implementation

[0064] The present invention will now be described in further detail.

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0066] The maintenance system, which forms the basis for the predictive maintenance method of the controller, will be introduced here first.

[0067] like Figure 1 As shown, a predictive maintenance system for controllers includes a controller and a display connected to it via CAN communication. The controller includes a data acquisition module, a data preprocessing module, a health assessment module, and an early warning decision module. The display is used to show data and provide visual alerts.

[0068] like Figure 2 As shown, a predictive maintenance method for controllers includes the following steps:

[0069] S1. Collect native data from the controller and ambient temperature;

[0070] The data acquisition module collects native data from the controller and ambient temperature in real time. ;

[0071] The controller's native data includes the internal power supply voltage. Chip temperature CPU utilization Memory usage Communication error frame rate Frequency of historical failures runtime The sampling frequency is 1 time per minute.

[0072] S2. Preprocess the collected data;

[0073] S21. The data preprocessing module calculates the average value for each data stream. and standard deviation Then adopt The principle is to iterate through all data points, and if the data... satisfy If the value is not found, it is considered an outlier and is removed to avoid interference from abnormal data caused by extreme operating conditions. At the same time, parameter fluctuations under normal operating conditions are retained to avoid data distortion caused by excessive removal. Linear interpolation is used to fill in the removed or missing data.

[0074] S22. Extract degradation-sensitive features from the cleaned data and calculate voltage fluctuations. Chip temperature rise Operating aging factors Failure rate .in, , , , ;

[0075] S23. Use the min-max normalization method to normalize the voltage fluctuation. Chip temperature rise Operating aging factors Failure rate CPU utilization Memory usage Communication error frame rate Once the data from each dimension is mapped to the [0,1] interval, the calculation formula is as follows: ,in, The original data, For the minimum value of data in this dimension, The maximum value of the data in this dimension. The normalized data corresponds to the voltage stability in subsequent steps. Chip temperature rise Failure rate CPU utilization Memory usage Communication error rate and operating aging factors By normalizing the data, the influence of different dimensions is eliminated, ensuring that data from each dimension can directly participate in subsequent weighted calculations and improving the accuracy of model evaluation.

[0076] S3. A weighted multi-index fusion algorithm is used to construct a controller health assessment model and calculate the current health of the controller.

[0077] S31. Determining the weights of indicators: Combining statistical data on engineering machinery failures, controller failure mechanisms, and expert experience, the weights of each evaluation indicator are determined. At the same time, the health threshold (ideal, corresponding to a health score of 100) and the failure threshold (worst, corresponding to a health score of 0) are clearly defined, as shown in Table (1). Table (1) ;

[0078] Table (1):

[0079]

[0080] S32. The health assessment module calculates the health score for each indicator and maps the normalized data of each indicator to a health score of 0 to 100 for each indicator.

[0081] Voltage stability health score ;

[0082] Chip temperature rise health score ;

[0083] Failure rate and health score ;

[0084] CPU utilization health score ;

[0085] Memory usage health score ;

[0086] Communication error rate health score ;

[0087] Running aging factor health points ;

[0088] S33. Weighted and integrated health scores from all individual indicators to obtain a comprehensive health score. , .

[0089] S4. Based on the comprehensive health score, classify the controller health level and output four-level warnings and corresponding maintenance suggestions;

[0090] Health level (80-100): The controller is operating normally, with no signs of aging, requiring no maintenance, and without any warnings.

[0091] Sub-health level (60-79): Level 1 warning, indicating slight aging, preventive maintenance is recommended;

[0092] Fault warning level (40-59): Level 2 alarm, obvious aging, key maintenance recommended;

[0093] Failure Level (0-39): Level 3 emergency alarm, there is a risk of sudden failure, it is recommended to stop the machine immediately for inspection and replacement of the controller;

[0094] The health score range for each level can be adjusted according to the actual situation.

[0095] S5. Warning information and corresponding maintenance suggestions are sent to the display via the CAN bus to achieve visual reminders;

[0096] Specifically, the early warning decision module encapsulates the information that needs to be warned according to the predefined CAN protocol and sends it to the display via the CAN bus. The display automatically retrieves the corresponding prompt interface according to the warning level and sends reminders to the user in the form of pop-ups, maintenance suggestion texts, etc., thereby realizing visual maintenance information notification.

[0097] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A predictive maintenance method for a controller, characterized in that, Includes the following steps: S1. Collect native data from the controller and ambient temperature; S2. Preprocess the collected data from each source, including outlier removal, missing value imputation, extraction of degradation-sensitive features, and normalization. S3. A weighted multi-index fusion algorithm is used to construct a controller health assessment model and calculate the current health of the controller. S4. Based on the comprehensive health score, classify the controller health level and output multi-level early warnings and corresponding maintenance suggestions.

2. The predictive maintenance method for controllers according to claim 1, characterized in that, In step S1, the controller's native data includes the internal power supply voltage. Chip temperature CPU utilization Memory usage Communication error frame rate Historical failure frequency runtime The sampling frequency is 1 time / minute; the ambient temperature .

3. The predictive maintenance method for controllers according to claim 1, characterized in that, In step S2, the specific methods of data preprocessing include: S21. Calculate the mean for each data stream. and standard deviation ,use In principle, if the data satisfy If the value is missing, it is identified as an outlier and removed. Linear interpolation is then used to fill in the missing or removed data. S22. Extract degradation-sensitive features from the cleaned data, including calculating voltage fluctuations. Chip temperature rise Operating aging factors Failure rate ;in, , , , ; S23. Using the min-max normalization method, the data of each dimension are mapped to the [0,1] interval. The calculation formula is as follows: .

4. The predictive maintenance method for controllers according to claim 1, characterized in that, In step S3, the specific steps for constructing the controller health assessment model include: S31. Based on the statistical data of engineering machinery failures, the failure mechanism of the controller and expert experience, determine the weight of each evaluation index, and define the health threshold and failure threshold. S32. Calculate the single-indicator health score and map the normalized data of each indicator to a single-indicator health score of 0 to 100. S33. Weighted and integrated health scores of all individual indicators are used to obtain the comprehensive health score H.

5. The predictive maintenance method for controllers according to claim 4, characterized in that, In step S32, the formula for calculating the health score of each individual indicator is as follows: Voltage stability health score ; in, The voltage stability data is after preprocessing. The healthy threshold for voltage stability, This is the failure threshold for voltage stability; Chip temperature rise health score ; in, This is the preprocessed chip temperature rise data. This refers to the chip's temperature rise health threshold. This is the chip temperature rise failure threshold; Failure rate and health score ; in, The failure rate, The failure threshold is the failure rate, and the health threshold is 0. CPU utilization health score ; in, For CPU utilization, The healthy threshold for CPU utilization. The failure threshold for CPU utilization; Memory usage health score ; in, For memory usage, This represents a healthy threshold for memory usage. This is the failure threshold for memory usage. Communication error rate health score ; in, For communication error rate, The failure threshold for communication error rate; Running aging factor health points ; in, The value of the aging factor after pretreatment.

6. The predictive maintenance method for controllers according to claim 4, characterized in that, In step S33, the formula for calculating the overall health score is as follows: ,in, As a weight for voltage stability, As a weight for chip temperature rise, To run the weights of the aging factor, As the weight of the failure rate, As a weight for CPU utilization, As a weight for memory usage, The weights are the communication error rate weights, and the sum of the weights is 1.

7. The predictive maintenance method for controllers according to claim 1, characterized in that, In step S4, the health level classification and corresponding warning levels are as follows: Health level 80-100: The controller is operating normally and there are no warnings. Sub-health level 60-79: Level 1 warning, preventive maintenance recommended; Fault warning level 40-59: Level 2 alarm, key maintenance recommended; Failure level 0-39: Level 3 emergency alarm, it is recommended to immediately stop the machine for inspection and replacement of the controller.

8. The predictive maintenance method for controllers according to claim 1, characterized in that, Also includes: S5. Send the warning information and corresponding maintenance suggestions to the display to achieve visual reminders.

9. A maintenance system, characterized in that, A controller for running the predictive maintenance method for a controller according to any one of claims 1-8, comprising a controller and a display communicatively connected thereto; The controller includes a data acquisition module, a data preprocessing module, a health assessment module, and an early warning decision module; The display is used to show data and provide visual reminders.

10. The maintenance system according to claim 9, characterized in that, The data acquisition module is used to collect the controller's native data and ambient temperature in real time.

11. The maintenance system according to claim 9, characterized in that, The data preprocessing module cleans each data stream, removes outliers, fills in missing values, extracts degradation-sensitive features from the cleaned data, and normalizes these degradation-sensitive feature data using min-max normalization.

12. The maintenance system according to claim 9, characterized in that, The health assessment module uses a weighted multi-index fusion algorithm to construct a controller health assessment model and calculate the controller's current health score.

13. The maintenance system according to claim 9, characterized in that, The early warning decision module classifies the controller's health level based on the comprehensive health score and outputs multi-level early warnings and corresponding maintenance suggestions.

14. The maintenance system according to claim 9, characterized in that, The early warning decision module encapsulates the information requiring early warning according to the predefined CAN protocol and sends it to the display via the CAN bus; The monitor automatically retrieves the corresponding prompt interface based on the warning level, and issues visual reminders to users through pop-ups and maintenance suggestion text.

Images