Device maintenance method, controller, and storage medium

Abstract

The application relates to the technical field of equipment maintenance, in particular to an equipment maintenance method, a controller and a storage medium, the method comprising the following steps: obtaining operation data of an electric equipment, wherein the operation data is collected by a multi-source sensor; performing feature engineering processing on the operation data to obtain multi-dimensional features; combining a time sequence model and a classification model to analyze the multi-dimensional features, so as to obtain the residual life and the fault category of the equipment; and generating a maintenance suggestion based on the residual life and the fault category. The technical scheme combines multi-source data to establish multi-dimensional features, combines a time sequence model and a classification model to analyze the multi-dimensional features, predicts performance attenuation and residual life, and identifies the fault category. Thus, the operation state of the equipment can be deeply understood, a maintenance suggestion can be generated, and a complete closed loop from data collection to decision support is formed. In this way, intelligent predictive real-time maintenance services are provided for the electric equipment, so that the operation reliability and operation and maintenance efficiency of the electric equipment are improved.

Description

Technical Field

[0001] This application relates to the field of equipment maintenance technology, and in particular to an equipment maintenance method, controller, and storage medium. Background Technology

[0002] With the increasing automation and intelligence of industrial production, electric equipment is playing an increasingly important role in modern industrial production. However, during long-term operation, these devices are inevitably affected by various factors, leading to performance degradation and malfunctions, which in turn affect production efficiency and safety.

[0003] Currently, equipment maintenance mainly relies on periodic maintenance and reactive repair. However, periodic maintenance often overlooks the actual operating condition of the equipment, which may lead to over-maintenance or untimely maintenance, while reactive repair can cause production interruptions and economic losses. Summary of the Invention

[0004] In view of this, this application provides a device maintenance method, controller and storage medium, which provide intelligent and predictive real-time maintenance services to improve the operational reliability and maintenance efficiency of electric equipment.

[0005] In a first aspect, this application provides an equipment maintenance method, comprising: acquiring operating data of an electric device, the operating data being collected by multiple source sensors; performing feature engineering processing on the operating data to obtain multi-dimensional features; analyzing the multi-dimensional features using a combination of time series models and classification models to obtain the remaining lifespan and fault categories of the device; and generating maintenance recommendations based on the remaining lifespan and fault categories.

[0006] In some embodiments, the aforementioned feature engineering process for the running data to obtain multidimensional features includes: cleaning the running data to remove outliers, fill in missing values, normalize or denoise; and establishing multidimensional features based on the cleaned running data.

[0007] In some embodiments, the aforementioned cleaning of the operating data includes: for normally distributed operating data, removing data points that are outside a preset range from the mean, wherein the preset range is greater than 2 standard deviations; for non-normally distributed operating data, calculating the local outlier factor and removing data points whose local outlier factor is greater than or equal to a preset threshold.

[0008] In some embodiments, the aforementioned cleaning of the running data includes: supplementing missing values ​​in the running data using interpolation.

[0009] In some embodiments, the interpolation process includes time-series domain-weighted interpolation.

[0010] In some embodiments, operating data includes speed, current, voltage, vibration, temperature, operating time, or battery health status.

[0011] In some embodiments, the time series model includes a long short-term memory model, and the classification model includes an XGBoost model.

[0012] In some embodiments, the fault categories include bearing wear, broken gear teeth, motor demagnetization, or battery aging.

[0013] Secondly, this application provides a controller, including: At least one processor, and A memory that is communicatively connected to at least one processor, wherein, The memory stores instructions that can be executed by at least one processor, which enables the at least one processor to perform the device maintenance method of the first aspect.

[0014] Thirdly, this application provides a computer-readable storage medium storing computer-executable instructions for causing a computer device to perform the device maintenance method as described in the first aspect.

[0015] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0016] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, which are not intended to limit the embodiments, and elements having the same reference numerals in the drawings are designated as similar elements.

[0017] Figure 1 This is a flowchart illustrating the equipment maintenance method in this application; Figure 2 This is a schematic diagram of the equipment maintenance system in this application; Figure 3 This is a schematic diagram of the controller in this application. Detailed Implementation

[0018] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used herein do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0021] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0022] Furthermore, the technical features involved in the various embodiments of this application described below can be combined with each other as long as they do not conflict with each other.

[0023] In modern industry, electric equipment such as fans, pumps, servo motors, and compressors play a vital role in production. However, during long-term operation, these devices are inevitably affected by various factors, leading to performance degradation or malfunctions, which in turn affect production efficiency and safety. Therefore, equipment maintenance is an important means to reduce malfunctions and ensure efficiency and safety.

[0024] Some equipment maintenance methods known to the inventors of this application are based on data-driven predictive maintenance, but they have many shortcomings. These methods typically lack an assessment of the health of the detection system itself and cannot effectively distinguish between genuine equipment failures and false alarms caused by sensor malfunctions. Furthermore, they struggle to track the slow degradation of equipment performance and are prone to prediction bias.

[0025] To address the aforementioned technical problems, this application provides an equipment maintenance method, controller, and storage medium. The method includes: acquiring operating data of an electric device, which is collected by multiple source sensors; performing feature engineering processing on the operating data to obtain multi-dimensional features; analyzing the multi-dimensional features using a combination of time-series and classification models to obtain the remaining lifespan and fault categories of the device; and generating maintenance recommendations based on the remaining lifespan and fault categories.

[0026] In this embodiment, operational data is collected from multiple sources of sensors. Fusing this multi-source data establishes multi-dimensional features, which, compared to single-dimensional features, facilitates comprehensive equipment monitoring and improves monitoring accuracy. A combined time-series model and a classification model are used to analyze these multi-dimensional features. The time-series model analyzes the changing trends of these features to predict performance degradation and remaining lifespan, while the classification model analyzes the feature distribution to identify fault categories. This allows for a deep understanding of the equipment's operational status, generating maintenance recommendations and forming a complete closed loop from data acquisition to decision support. Thus, intelligent and predictive real-time maintenance services are provided for electric equipment, improving its operational reliability and maintenance efficiency.

[0027] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0028] Please see Figure 1 , Figure 1 This is a flowchart illustrating a device maintenance method in some embodiments of this application. The method can be executed by a controller, which can be applied to electric equipment to perform intelligent, predictive, and real-time maintenance.

[0029] The above method S100 includes the following steps: S10: Acquire the operating data of the electric equipment, which is collected by multi-source sensors.

[0030] Understandably, during operation, the running status of electric equipment can be monitored by sensors, thereby collecting operational data. These sensors may include current sensors, voltage sensors, speed sensors, accelerometers, temperature sensors, or timing units, etc.

[0031] In some embodiments, the aforementioned operating data includes rotational speed, current, voltage, vibration, temperature, operating time, or battery health status. The battery health status can be calculated by the battery management system based on sensor data such as voltage, current, and temperature, and the battery management system sends the battery health status to the controller. Other sensors send their collected data to the controller, thus allowing the controller to acquire the aforementioned operating data.

[0032] S20: Perform feature engineering on the running data to obtain multidimensional features.

[0033] Feature engineering refers to extracting structured features from operational data that are beneficial for model learning. For example, a set of structured feature vectors can be extracted from multiple dimensions of operational data, such as time domain, frequency domain, and statistics. Each dimension represents a quantitative indicator of the device's operating status, thus obtaining multidimensional features.

[0034] In some embodiments, the aforementioned step S20 specifically includes: S21: Clean the running data to remove outliers, fill in missing values, normalize or denoise.

[0035] Understandably, the original operational data has flaws that can interfere with model analysis. Therefore, the operational data needs to be cleaned; specifically, outliers are removed, missing values ​​are filled, and then normalization or noise reduction is performed.

[0036] S22: Establish multidimensional features based on the cleaned operational data.

[0037] Extracting multidimensional features from cleaned operational data can eliminate the negative impacts of outliers, missing values, noise, and units of measurement, thereby avoiding feature distortion and ensuring that features are true, stable, and complete, which in turn helps improve the accuracy of subsequent model recognition.

[0038] In some embodiments, the "cleaning of runtime data" in step S21 includes: S211: For normally distributed operating data, remove data points that are outside the preset range of the mean, where the preset range is greater than 2 standard deviations.

[0039] For electric equipment, operating data such as current, voltage, temperature, and steady-state speed follow a normal distribution. Understandably, normally distributed data exhibits a bell-shaped curve, symmetrical around the mean u, with the data volume gradually decreasing on both sides. Therefore, data points far from the mean, located outside a certain range, contain significant errors. Thus, removing data points outside the preset range of the mean effectively eliminates interfering data points.

[0040] Optionally, the preset range is greater than 2 standard deviations, for example, ±3σ. This removes data points that exceed the range u±3σ, effectively eliminating outliers.

[0041] S212: For non-normally distributed running data, calculate the local outlier factor and remove data points whose local outlier factor is greater than or equal to a preset threshold.

[0042] Understandably, operational data such as vibration data, battery health status, cumulative operating time, and transient current and transient speed under equipment start-up, shutdown, and variable load conditions are affected by mechanical shock, load switching, battery aging, and dynamic changes in operating conditions, causing their distribution to deviate from a normal distribution. For this type of operational data, a local outlier factor is calculated, and data points with a local outlier factor greater than or equal to a preset threshold are identified as outliers and removed.

[0043] The Local Outlier Factor (LOF) is a density-based anomaly detection metric used to measure the difference in local density between a single data point and its neighboring points. If the local density of a point is significantly lower than that of its neighborhood, its LOF value will be significantly higher, and it will be identified as an outlier. The closer the LOF is to 1, the more consistent the density of the data point with its surrounding data, indicating that it is normal data. For example, if the preset threshold is 2.0, then data points with an LOF ≥ 2.0 will be removed as outliers.

[0044] In this embodiment, for normally distributed data, the standard deviation criterion is used to remove outliers; for non-normal data, the local outlier factor (LOF) is used to remove outliers. The two methods complement each other and can comprehensively ensure data quality.

[0045] In some embodiments, the "cleaning of runtime data" in step S21 includes: S213: For missing values ​​in the running data, interpolation is used to supplement them.

[0046] It is understandable that interpolation can be used to supplement missing values, thereby repairing breakpoints in time-series data, ensuring data integrity and continuity, effectively avoiding feature calculation anomalies, reducing data loss, and improving the reliability of subsequent model analysis and fault detection.

[0047] In some embodiments, the aforementioned interpolation process includes time-series domain-weighted interpolation.

[0048] In the temporal domain weighted interpolation, missing points are determined based on nearby known points. The closer a known point is to the missing point, the greater its weight, and the farther a known point is from the missing point, the smaller its weight.

[0049] For example, for missing values ​​at time t The following formula is used for calculation:

[0050] in, For the nearest valid time, Here, is the decay time constant, i is the index of a known neighboring point, and k is the number of known neighboring points. The weights of the nearest known points, It is the i-th nearest valid data point, at time... The corresponding runtime data values.

[0051] Thus, by using time-series domain-weighted interpolation, missing data can be accurately compensated.

[0052] S30: Use a combination of time series model and classification model to analyze multidimensional features and obtain the remaining life and failure category of the equipment.

[0053] For example, the multidimensional features include 48 dimensions, which are input into both the time-series model and the classification model. The time-series model analyzes the changing trends of the 48 dimensions over time and outputs the remaining lifespan of the device. The classification model extracts statistical feature distribution and time-frequency domain features based on feature engineering to identify the fault category.

[0054] In some embodiments, the time-series model includes a long short-term memory model.

[0055] For example, the structure of a long short-term memory model includes: Input layer, Window length ; Two LSTM layers, each with 64 units, Dropout=0.3; The fully connected layer has 32 neurons and uses the ReLU activation function. The output layer consists of one neuron, linearly activated, that outputs the RUL prediction value.

[0056] By employing an LSTM model with the aforementioned structure, it is possible to effectively learn the dynamic changes of 48-dimensional high-dimensional features over time, capture the trend of equipment performance degradation, and improve the prediction accuracy of remaining lifespan. Furthermore, this LSTM model exhibits strong robustness, is adaptable to complex industrial conditions, and can achieve accurate prediction of equipment lifespan.

[0057] In some embodiments, the time series model may also be a GRU, Transformer, or TCN model.

[0058] In some embodiments, the classification model includes an XGBoost model. This XGBoost model consists of multiple decision trees connected in series. It continuously trains new trees to correct the prediction errors of preceding models, and finally sums the results of all trees to obtain the final output. In some implementations, the number of trees is 150, the maximum depth is 6, the learning rate is 0.05, the subsampling ratio is 0.8, and the column sampling ratio is 0.7.

[0059] In some embodiments, fault categories include bearing wear, gear tooth breakage, motor demagnetization, or battery aging. Thus, the XGBoost model outputs a probability distribution for bearing wear, gear tooth breakage, motor demagnetization, or battery aging. The fault category with the highest probability is taken as the output.

[0060] In some embodiments, the classification model may also include LightGBM, CatBoost, or a multilayer perceptron.

[0061] S40: Generate maintenance recommendations based on remaining lifespan and fault type.

[0062] For example, maintenance recommendations include maintenance time and maintenance content. Taking into account both the remaining lifespan of the equipment and the type of failure helps to generate a better maintenance strategy compared to considering a single factor.

[0063] In summary, this application integrates multi-source data to establish multi-dimensional features. Compared to single-dimensional features, this approach facilitates comprehensive equipment monitoring and improves monitoring accuracy. By combining time-series and classification models to analyze these multi-dimensional features, the time-series model analyzes the changing trends of these features to predict performance degradation and remaining lifespan, while the classification model analyzes the feature distribution of these features to identify fault categories. This enables a deeper understanding of the equipment's operating status, generating maintenance recommendations and forming a complete closed loop from data acquisition to decision support. Consequently, this provides intelligent, predictive, and real-time maintenance services for electric equipment, improving its operational reliability and maintenance efficiency.

[0064] Please see Figure 2 This application also provides an equipment maintenance system 200, including a data acquisition layer 201, a data preprocessing layer 202, a model training layer 203, a model application layer 204, and a decision optimization layer 205.

[0065] The data acquisition layer is responsible for collecting operational data from the electric equipment, such as speed, current, voltage, vibration, temperature, operating time, and battery health. Based on multiple sensors, the data acquisition layer comprehensively monitors the equipment's operating status and collects operational data.

[0066] The data preprocessing layer cleans the collected operational data, including removing outliers, filling in missing values, and performing normalization or noise reduction. This ensures the consistency and accuracy of the operational data. Furthermore, a Kalman filter can be used to eliminate noise in the operational data, improving its signal-to-noise ratio.

[0067] In some implementations, for normally distributed data, data points outside a preset range (greater than two standard deviations) from the mean are removed; for non-normally distributed data, local outlier factors are calculated, and data points with local outlier factors greater than or equal to a preset threshold are removed. In this way, the two methods complement each other, comprehensively ensuring data quality.

[0068] In some implementations, time-series domain-weighted interpolation is used to supplement missing values ​​in the runtime data, which can accurately compensate for missing data.

[0069] The model training layer stores the constructed model framework, which includes a time-series model and a classification model. The time-series model analyzes the changing trends of runtime data over time, predicting performance degradation and remaining lifetime. The classification model extracts statistical feature distributions and time-frequency domain features based on feature engineering to identify fault categories. For example, it identifies bearing wear, gear tooth breakage, motor demagnetization, or battery aging.

[0070] The model application layer calls the time series model and classification model in the above model training layer to monitor the electric equipment in real time based on the operating data, predict the remaining life and identify the fault category, and dynamically update the prediction and identification results.

[0071] The decision optimization layer comprehensively considers multiple dimensions such as the equipment's remaining lifespan (health status, deterioration trend) and the confidence level of predicted failure categories to automatically generate maintenance recommendations. This helps users formulate better maintenance strategies.

[0072] Thus, the entire equipment maintenance system adopts a layered and modular design, with each layer communicating through standard interfaces, enabling efficient data transmission and information sharing. In some implementations, the system also supports continuous learning capabilities; the time-series and classification models in the model training layer can self-train based on runtime data, fine-tuning model parameters.

[0073] This application also provides a control device; please refer to [link / reference]. Figure 3 The controller 300 includes a processor 301 and a memory 302. The processor 301 is connected to the memory 302, for example, the processor 301 can be connected to the memory 302 via a bus.

[0074] Processor 301 is configured to support the control device in executing the steps of the device maintenance method described in the above embodiments. Processor 301 may be a central processing unit (CPU), a network processor (NP), a hardware chip, or any combination thereof. The aforementioned hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0075] Memory 302 is used to store program code, etc. Memory 302 may include volatile memory (VM), such as random access memory (RAM); memory 302 may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory 302 may also include combinations of the above types of memory.

[0076] In some possible scenarios, the processor 301 may call the program code to perform the following operations: acquire operating data of the electric equipment, which is collected by multiple source sensors; perform feature engineering on the operating data to obtain multi-dimensional features; analyze the multi-dimensional features using a combination of time series model and classification model to obtain the remaining lifespan and fault category of the equipment; and generate maintenance recommendations based on the remaining lifespan and fault category.

[0077] In some embodiments, the aforementioned feature engineering process for the running data to obtain multidimensional features includes: cleaning the running data to remove outliers, fill in missing values, normalize or denoise; and establishing multidimensional features based on the cleaned running data.

[0078] In some embodiments, the aforementioned cleaning of the operating data includes: for normally distributed operating data, removing data points that are outside a preset range from the mean, wherein the preset range is greater than 2 standard deviations; for non-normally distributed operating data, calculating the local outlier factor and removing data points whose local outlier factor is greater than or equal to a preset threshold.

[0079] In some embodiments, the aforementioned cleaning of the running data includes: supplementing missing values ​​in the running data using interpolation.

[0080] In some embodiments, the interpolation process includes time-series domain-weighted interpolation.

[0081] In some embodiments, operating data includes speed, current, voltage, vibration, temperature, operating time, or battery health status.

[0082] In some embodiments, the time series model includes a long short-term memory model, and the classification model includes an XGBoost model.

[0083] In some embodiments, the fault categories include bearing wear, broken gear teeth, motor demagnetization, or battery aging.

[0084] Based on the same inventive concept, the principle and beneficial effects of the controller provided in the embodiments of this application are similar to the principle and beneficial effects of the method embodiments of this application. Please refer to the principle and beneficial effects of the method embodiments. For the sake of brevity, they will not be repeated here.

[0085] This application also provides an electric device, including the controller described in the above embodiments. This electric device can be a fan, water pump, servo motor, compressor, or other power tool. This enables the electric device to have intelligent maintenance functions.

[0086] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the device maintenance method as described in the foregoing embodiments.

[0087] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A method for maintaining equipment, characterized in that, include: The operating data of the electric equipment is acquired by multi-source sensors. The operational data is subjected to feature engineering to obtain multidimensional features; The remaining lifespan and fault category of the equipment are obtained by combining time-series and classification models to analyze the multidimensional features. Maintenance recommendations are generated based on the remaining lifespan and the fault category.

2. The method according to claim 1, characterized in that, The runtime data is subjected to feature engineering to obtain multidimensional features, including: The running data is cleaned to remove outliers, fill in missing values, normalize, or denoise. The multidimensional features are established based on the cleaned operational data.

3. The method according to claim 2, characterized in that, The cleaning of the operational data includes: For normally distributed operating data, data points that are outside the preset range of the mean are removed, where the preset range is greater than 2 standard deviations. For non-normally distributed operating data, calculate the local outlier factor and remove data points whose local outlier factor is greater than or equal to a preset threshold.

4. The method according to claim 2, characterized in that, The cleaning of the operational data includes: For missing values ​​in the runtime data, interpolation is used to supplement them.

5. The method according to claim 4, characterized in that, The interpolation processing method includes time-series domain-weighted interpolation.

6. The method according to claim 1, characterized in that, The operating data includes speed, current, voltage, vibration, temperature, operating time, or battery health status.

7. The method according to claim 1, characterized in that, The time series model includes a long short-term memory model, and the classification model includes an XGBoost model.

8. The method according to claim 1, characterized in that, The fault categories include bearing wear, broken gear teeth, motor demagnetization, or battery aging.

9. A controller, characterized in that, include: At least one processor, and The memory communicatively connected to the at least one processor, wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer device to perform the method as described in any one of claims 1-8.

Images