Predictive maintenance optimization method, arrangement, device and engineering machinery

Abstract

The application relates to the technical field of fault prediction, and provides a predictive maintenance optimization method, device, equipment and engineering machinery. The method comprises the following steps: acquiring a first time point of occurrence of a target fault predicted by a fault prediction model and a second time point of actual occurrence of the target fault; comparing a time difference value between the first time point and the second time point with an effective duration corresponding to the target fault to determine an accurate prediction fault number of the time difference value being less than the effective duration, the accurate prediction fault number being used for calculating an evaluation index of the fault prediction model, and the fault prediction model being optimized according to the evaluation index. The effective duration gives the time point of occurrence of the fault predicted by the fault prediction model with timeliness, the fault tolerance space of the predicted time point of occurrence of the fault is improved, the evaluation index of the fault prediction model calculated by the accurate prediction fault number obtained in this way is more reasonable and accurate, the best model parameter can be obtained, and therefore the prediction effect of the fault prediction model is improved.

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, and in particular to a predictive maintenance optimization method, apparatus, equipment, and engineering machinery. Background Technology

[0002] Predictive maintenance, due to its high timeliness in fault detection, is widely used in industries such as industrial equipment and commercial vehicles. Currently, predictive maintenance mainly includes two approaches: mechanistic model-based fault prediction and mathematical model-based fault prediction. Mechanism-based prediction relies on the experience accumulated by R&D personnel during the R&D process or in the industry, resulting in a comprehensive understanding of component faults and a clear identification of relevant parameters and their trends during the fault occurrence process. Mathematical model-based prediction primarily uses big data models to analyze various status data, operating conditions, and actual faults of the equipment to identify correlations between data, parameters, and logic. Both types of prediction methods are ultimately evaluated for model effectiveness and usability based on accuracy. Existing evaluation methods that rely on a one-to-one comparison between predicted and actual faults fail to consider the timeliness of fault prediction, leading to inaccurate evaluation results for the fault prediction model. Consequently, the optimized model parameters based on the evaluation results are not optimal, impacting the model's predictive performance. Summary of the Invention

[0003] This invention provides a predictive maintenance optimization method, apparatus, equipment, and engineering machinery to address the shortcomings of existing technologies, such as inaccurate evaluation results of fault prediction models and the fact that the optimized model parameters obtained through evaluation results are not optimal, thus affecting the model's predictive performance.

[0004] This invention provides a predictive maintenance optimization method, comprising:

[0005] Obtain the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs;

[0006] The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to determine the number of accurate predicted faults when the time difference is less than the effective duration.

[0007] Based on the accurate prediction of the number of failures, the evaluation index of the failure prediction model is calculated, and the failure prediction model is optimized according to the evaluation index.

[0008] According to the predictive maintenance optimization method provided by the present invention, the evaluation index includes time accuracy; the step of calculating the evaluation index of the fault prediction model based on the accurately predicted number of faults includes:

[0009] Based on the target time point of each accurate prediction of the target fault in the number of accurate predicted faults, and the second time point of the actual occurrence of the target fault corresponding to the target time point, the prediction duration of the target fault by the fault prediction model is calculated.

[0010] Calculate the time standard deviation of the fault prediction model based on the predicted duration, the effective duration, and the number of accurately predicted faults.

[0011] The time accuracy of the fault prediction model is obtained by calculating the ratio of the target difference to the effective duration; the target difference is the difference between the effective duration and the time standard deviation.

[0012] According to the predictive maintenance optimization method provided by the present invention, the step of comparing the time difference between the first time point and the second time point with the effective duration corresponding to the target fault to determine the accurate number of predicted faults where the time difference is less than the effective duration includes:

[0013] The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to obtain the number of effective predicted faults where the predicted time point corresponding to the first time point is before the second time point, and based on the number of effective predicted faults, the number of accurate predicted faults where the time difference is less than the effective duration is determined.

[0014] According to the predictive maintenance optimization method provided by the present invention, the evaluation index includes prediction accuracy; the step of calculating the evaluation index of the fault prediction model based on the accurately predicted number of faults includes:

[0015] Obtain the union of the number of effective predicted faults and the number of actual occurrences of the target fault;

[0016] The prediction accuracy of the fault prediction model is calculated based on the ratio of the number of accurately predicted faults to the number of faults in the union set.

[0017] According to the predictive maintenance optimization method provided by the present invention, before comparing the time difference between the first time point and the second time point with the effective duration corresponding to the target fault, the method further includes:

[0018] Obtain the fault type of the target fault;

[0019] Based on the fault type, determine the effective duration corresponding to the target fault.

[0020] According to the predictive maintenance optimization method provided by the present invention, the effective duration includes equipment runtime and natural duration.

[0021] According to the predictive maintenance optimization method provided by the present invention, the effective duration varies with gradient, and the optimization of the fault prediction model based on the evaluation index includes:

[0022] During the training process of the fault prediction model, the model parameters of the fault prediction model are adjusted according to the evaluation index to optimize the fault prediction model.

[0023] The effective duration is adjusted based on a preset gradient value. The steps of obtaining the first time point at which the target fault is predicted by the fault prediction model and the second time point at which the target fault actually occurs are returned and executed until the effective duration reaches a first preset threshold and the evaluation index reaches a second preset threshold.

[0024] The present invention also provides a predictive maintenance optimization device, comprising:

[0025] The data acquisition module is used to acquire the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs.

[0026] The fault screening module is used to compare the time difference between the first time point and the second time point with the effective duration corresponding to the target fault, so as to determine the accurate number of predicted faults when the time difference is less than the effective duration.

[0027] The model optimization module is used to calculate the evaluation index of the fault prediction model based on the number of accurately predicted faults, and to optimize the fault prediction model according to the evaluation index.

[0028] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the predictive maintenance optimization methods described above.

[0029] The present invention also provides an engineering machinery, including an engineering machinery body, wherein the engineering machinery body is provided with a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein when the program or instructions are executed by the processor, the steps of any of the above-described predictive maintenance optimization methods are implemented.

[0030] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the predictive maintenance optimization method as described above.

[0031] The predictive maintenance optimization method, apparatus, equipment, and engineering machinery provided by this invention obtains the first time point of the target fault predicted by the fault prediction model and the second time point of the actual target fault occurrence; compares the time difference between the first and second time points with the effective duration corresponding to the target fault, and determines the accurate number of predicted faults where the time difference is less than the effective duration; and optimizes the fault prediction model by calculating an evaluation index based on this accurate number of predicted faults. By using the effective duration, the predicted fault occurrence time point of the fault prediction model is given timeliness, increasing the fault tolerance space of the predicted fault occurrence time point. The evaluation index of the fault prediction model calculated from the accurate number of predicted faults obtained in this way is more reasonable and accurate, which is conducive to obtaining the optimal model parameters, thereby improving the prediction effect of the fault prediction model. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0033] Figure 1 This is a flowchart illustrating the predictive maintenance optimization method provided by the present invention;

[0034] Figure 2 This is a schematic diagram of the prediction accuracy provided in an embodiment of the present invention;

[0035] Figure 3 This is a schematic diagram of the predictive maintenance optimization device provided by the present invention;

[0036] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0038] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, as a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0039] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0040] This invention provides a predictive maintenance optimization method for optimizing the fault prediction model used in predictive maintenance. By introducing the concept of effective duration to evaluate the fault prediction model, it gives fault prediction timeliness, which helps to improve the evaluation accuracy of the fault prediction model, thereby improving the model optimization effect, making the model parameters optimal, and thus improving the prediction effect of the fault prediction model.

[0041] Specifically, refer to Figure 1 , Figure 1 This is a flowchart illustrating the predictive maintenance optimization method provided in an embodiment of the present invention, based on... Figure 1 The predictive maintenance optimization method provided in this embodiment of the invention includes:

[0042] Step 100: Obtain the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs;

[0043] First, the first time point at which the target fault occurs, as predicted by the fault prediction model, and the second time point at which the target fault actually occurs are obtained. The second time point at which the target fault actually occurs is the actual time of occurrence of the target fault as indicated in the training samples of the fault prediction model. The first time point at which the target fault occurs, as predicted by the fault prediction model, can be obtained after the fault prediction model has been trained or during the training process.

[0044] Preferably, the fault prediction model can be a mechanistic model or a mathematical model, without specific limitations. Regarding the target fault, it is known that for the same equipment or vehicle, different components correspond to different types of faults, and the same component may also correspond to one or more types of faults. Therefore, the fault prediction model can be a multi-classification model. When training the fault prediction model, training samples containing multiple faults can be used, and the target fault can be any one of these faults.

[0045] Step 200: Compare the time difference between the first time point and the second time point with the effective duration corresponding to the target fault to determine the accurate number of predicted faults where the time difference is less than the effective duration;

[0046] Based on the first time point predicted by the acquired fault prediction model and the second time point where the target fault actually occurs, the time difference between the first and second time points is compared with the effective duration corresponding to the target fault, thereby determining the accurate number of predicted faults where the time difference is less than the effective duration.

[0047] It is understandable that at the first time point predicted by the fault prediction model as the occurrence of the target fault, the target fault may or may not actually occur. Therefore, there is a certain correspondence between the first time point predicted by the fault prediction model and the second time point in time when the target fault actually occurs, but it is not a one-to-one correspondence. In fact, at the first time point predicted by the fault prediction model as the occurrence of the target fault, the target fault may or may not actually occur, and the target fault may also actually occur within a period of time before or after the first time point.

[0048] In existing technical solutions, the evaluation method for fault prediction models involves comparing the predicted fault occurrence time with the actual fault occurrence time. Only when the two match is the fault prediction model considered to have made an accurate prediction. However, in reality, the fault prediction model's prediction of the fault occurrence time inherently has a certain degree of anticipation. The actual fault occurrence time does not necessarily have to be exactly the same as the predicted time. Whether before or after the predicted time, as long as the difference between the predicted time and the actual fault occurrence time is within an acceptable range, the fault prediction model can be considered to have accurately predicted the fault occurrence time. Therefore, introducing an effective duration, giving the fault prediction model's prediction results a timeliness, helps improve the accuracy of the evaluation results of the fault prediction model, thereby improving the optimization effect of the fault prediction model and facilitating the acquisition of the optimal parameters of the fault prediction model. Preferably, the effective duration limits the acceptable range of the difference between the predicted time and the actual fault occurrence time.

[0049] Step 300: Based on the accurate prediction of the number of faults, calculate the evaluation index of the fault prediction model, and optimize the fault prediction model according to the evaluation index.

[0050] Based on the accurate prediction of the number of failures, an evaluation index for the failure prediction model is calculated. This evaluation index includes at least the prediction accuracy. The failure prediction model is then optimized based on this evaluation index to improve its performance in predicting failures, thereby optimizing predictive maintenance.

[0051] It is evident that, based on the effective duration, the fault prediction model provides a certain time tolerance for the predicted fault occurrence time, giving the predicted fault occurrence time a certain timeliness. This makes the evaluation indicators of the fault prediction model more reasonable and accurate, which is conducive to obtaining the best model parameters and improving the prediction effect of the fault prediction model.

[0052] In this embodiment, the first time point at which the target fault occurs, predicted by the fault prediction model, and the second time point at which the target fault actually occurs are obtained. The time difference between the first and second time points is compared with the effective duration corresponding to the target fault to determine the number of accurately predicted faults where the time difference is less than the effective duration. Based on this number of accurately predicted faults, an evaluation index for the fault prediction model is calculated to optimize the model. By using the effective duration, the predicted fault occurrence time point of the fault prediction model is given timeliness, increasing the fault tolerance space of the predicted fault occurrence time point. The evaluation index of the fault prediction model calculated from the number of accurately predicted faults obtained in this way is more reasonable and accurate, which is conducive to obtaining the optimal model parameters and thus improving the prediction effect of the fault prediction model.

[0053] Preferably, the evaluation metrics for the fault prediction model include time accuracy and prediction accuracy. When time accuracy is included as an evaluation metric, step 300, based on the number of accurately predicted faults, calculates the evaluation metrics for the fault prediction model, which further includes:

[0054] Step 310: Based on the target time point of each accurate prediction of the target fault occurrence in the number of accurate predicted faults, and the second time point of the actual occurrence of the target fault corresponding to the target time point, calculate the prediction duration of the target fault by the fault prediction model.

[0055] Step 320: Calculate the time standard deviation of the fault prediction model based on the predicted duration, the effective duration, and the number of accurately predicted faults;

[0056] Step 330: Calculate the ratio of the target difference to the effective time to obtain the time accuracy of the fault prediction model; the target difference is the difference between the effective duration and the time standard deviation.

[0057] When calculating the time accuracy of the fault prediction model, the time standard deviation of the model is first calculated based on the effective duration. Then, the time accuracy of the model is calculated based on this time standard deviation and the effective duration. Specifically, when calculating the time standard deviation, the prediction duration of the model is calculated based on the target time point of each accurate prediction of the target fault in the number of accurate fault predictions, and the second time point of the actual occurrence of the target fault corresponding to the target time point. In other words, how far in advance the model actually predicted the occurrence of the target fault is. It should be noted that for each accurate prediction of the target fault, there is a corresponding actual time point of occurrence of the target fault.

[0058] Furthermore, the time standard deviation of the fault prediction model is calculated based on the prediction duration, effective duration, and the number of accurately predicted faults. Preferably, the standard deviation between each prediction time point and the actual occurrence time point is as shown in Formula 1 below:

[0059]

[0060] Where n represents the number of accurately predicted failures.

[0061] Furthermore, the difference between the effective duration and the time standard deviation is used as the target difference. The ratio of the target difference to the effective duration is calculated to obtain the time accuracy of the fault prediction model, as shown in Formula 2 below:

[0062]

[0063] Preferably, in step 200, the time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to determine the accurate number of predicted faults where the time difference is less than the effective duration. Specifically, this includes:

[0064] Step 210: Compare the time difference between the first time point and the second time point with the effective duration corresponding to the target fault to obtain the number of effective predicted faults where the predicted time point corresponding to the first time point is before the second time point, and determine the number of accurate predicted faults where the time difference is less than the effective duration based on the number of effective predicted faults.

[0065] The time difference between the first and second time points is compared with the effective duration corresponding to the target fault to obtain the number of valid predicted faults before the second time point corresponding to the prediction time point of the first time point, and the number of accurate predicted faults with a time difference less than the effective duration based on the number of valid predicted faults. It can be understood that in the prediction results for the target fault, based on the second time point where the target fault actually occurs, the prediction results can be divided into valid predictions and invalid predictions. Valid predictions are those where the prediction time point corresponding to the first time point is before the second time point where the target fault actually occurs, while predictions where the prediction time point is after the second time point where the target fault actually occurs are invalid predictions. The prediction time point is the time point at which the fault prediction model generates the prediction result, that is, the time point at which the first time point is generated. Predictions of the target fault generated before the actual occurrence of the target fault are valid predictions, while predictions generated after the actual occurrence of the target fault are invalid predictions. Furthermore, based on the number of valid predicted faults, the number of accurate predicted faults with a time difference between the first and second time points less than the effective duration is determined.

[0066] Preferably, when the evaluation metric for the fault prediction model includes prediction accuracy, step 300, which calculates the evaluation metric for the fault prediction model based on the number of accurately predicted faults, may further include:

[0067] Step 340: Obtain the union of the number of effective predicted faults and the actual number of occurrences of the target fault;

[0068] Step 350: Calculate the prediction accuracy of the fault prediction model based on the ratio of the number of accurately predicted faults to the number of faults in the union set.

[0069] First, obtain the union of the number of effectively predicted faults and the number of actual occurrences of the target fault; that is, the union of effectively predicted faults and actual occurrences. Then, calculate the prediction accuracy of the fault prediction model based on the ratio of the number of accurately predicted faults to the number of faults in the union. The reason for taking the union of the number of effectively predicted faults and the number of actual occurrences is that the fault prediction model's prediction results for actual occurrences include: the target fault actually occurred, and the fault prediction model accurately predicted its occurrence; the target fault actually occurred, but the fault prediction model did not accurately predict its occurrence; the fault prediction model predicted the target fault would occur, but the target fault did not actually occur; and the fault prediction model did not predict the target fault would occur, and the target fault did not actually occur. Conversely, the fault prediction model's accurate prediction results for the target fault include: the target fault actually occurred, and the fault prediction model accurately predicted its occurrence; and the fault prediction model did not predict the target fault would occur, and the target fault did not actually occur.

[0070] By taking the union of the number of effectively predicted faults and the number of actual faults, we can obtain all the prediction results of the fault prediction model for the target fault. Based on this, the ratio of the number of accurately predicted faults to the number of faults in the union is calculated, thus yielding the prediction accuracy of the fault prediction model. Preferably, the formula for calculating the accuracy of the fault prediction model is as follows:

[0071]

[0072] Preferably, in step 200, before comparing the time difference between the first time point and the second time point with the effective duration corresponding to the target fault, the following may be included:

[0073] Step 201: Obtain the fault type of the target fault;

[0074] Step 202: Determine the effective duration corresponding to the target fault based on the fault type.

[0075] Before comparing the time difference between the first and second time points with the effective duration corresponding to the target fault, the effective duration corresponding to the target fault must first be obtained. This effective duration is determined based on the fault type of the target fault. Specifically, the fault type of the target fault is first obtained, and the effective duration corresponding to the target fault is determined based on this fault type. Different types of faults correspond to different effective durations, which can be user-configurable. For example, a user can configure the effective duration for a fault type by specifying how far in advance a user expects a certain fault to be predicted. For fault types where the user has not configured an effective duration, the effective duration can be adaptively determined based on accumulated empirical data. No specific limitations are placed on the method for determining the effective duration here.

[0076] Furthermore, it should be noted that the effective duration includes two different calculation methods: equipment runtime and natural duration. Some faults are related to the operating state of the equipment and are faults that occur as the equipment operates. For state values ​​that may only be consumed when the equipment is running or when the parts are used, the calculation is based on the cumulative equipment runtime. For example, when urea is consumed only when the vehicle is running, the effective duration should be calculated based on the vehicle's runtime. On the other hand, some faults are not related to the operating state of the equipment but are faults that occur with natural time. For state values ​​that gradually change according to natural time, such as battery charge, the effective duration should be calculated based on natural time.

[0077] Current predictive maintenance methods lack flexible calculations for different fault types; instead of defining accurate predictions, a fixed and uniform calculation method is used. This makes the evaluation of the fault prediction model inaccurate for both high-speed and slow-onset faults. However, by introducing the concept of effective duration, setting different effective durations for different types of faults, and establishing different calculation methods for these effective durations, a reasonable definition of accurate predictions is established. This leads to a more reasonable and accurate evaluation of the fault prediction model and facilitates the optimization of its parameters.

[0078] Preferably, refer to Figure 2 The diagram illustrates the prediction accuracy of the fault prediction model. For example, using heavy-duty trucks and other commercial vehicles as the fault prediction target, in... Figure 2 The actual meanings of each parameter are as follows:

[0079] Actual fault: The fault reported by the same vehicle has a fault status value from open to close. The same fault reported during this period is considered as the same fault. The vehicle reports the fault in the form of fault code. Different faults have unique fault codes.

[0080] Fault prediction: The fault prediction model predicts faults based on the data reported by the vehicle. For the same vehicle, the same fault predicted during the period from prediction to the occurrence of the fault to the failure to predict the fault is considered as the same fault.

[0081] Effective time point: How far in advance the user believes the fault should be predicted, or how far in advance data based on accumulated experience indicates the fault should be predicted and resolved, is the effective timeframe for the fault. This effective timeframe corresponds to the point in time before the actual fault occurs. For different types of faults, the effective timeframe includes two different calculation methods: vehicle running time and natural time.

[0082] Effective prediction: A failure that is predicted before it actually occurs is considered an effective prediction;

[0083] Invalid prediction: A fault predicted after the fault has already occurred and a fault code has been reported is an invalid prediction;

[0084] Accurate prediction: In effective prediction, the time difference between the predicted time of failure and the actual time of failure is less than the effective prediction duration.

[0085] Incorrect prediction: In effective prediction, the time difference between the predicted time of failure and the actual time of failure is greater than or equal to the effective prediction duration.

[0086] In one embodiment, accurate prediction and incorrect prediction are determined by the expiration of the predicted fault. Specifically, starting from the time point when the prediction result of the fault prediction model is generated, if the vehicle reports the fault code of the fault predicted by the fault prediction model within twice the effective time period, then the predicted fault has expired and is determined to be an accurate prediction by the fault prediction model; if the vehicle does not report the fault code of the fault predicted by the fault prediction model when twice the effective time period has been reached, then the predicted fault has expired and is determined to be an incorrect prediction.

[0087] Preferably, during the training process of the fault prediction model, the effective duration varies with gradients. In step 300, optimizing the fault prediction model based on evaluation metrics further includes:

[0088] Step 301: During the training process of the fault prediction model, the model parameters of the fault prediction model are adjusted according to the evaluation index to optimize the fault prediction model.

[0089] Step 302: Adjust the effective duration based on the preset gradient value, return and execute the steps of obtaining the first time point of the target fault predicted by the fault prediction model and the second time point of the actual occurrence of the target fault, until the effective duration reaches the first preset threshold and the evaluation index reaches the second preset threshold.

[0090] When optimizing the fault prediction model, firstly, during the training process, the model parameters are adjusted according to the evaluation metrics to optimize the model. Then, based on preset gradient values, the effective duration corresponding to each target fault is adjusted. Step 100 is then executed again to recalculate the evaluation metrics of the fault prediction model until the effective duration reaches a first preset threshold and the evaluation metrics reach a second preset threshold. At this point, the training of the fault prediction model is complete, resulting in a fault prediction model for predictive maintenance.

[0091] The first preset threshold corresponding to the effective duration can be user-defined, aiming to predict the duration of a fault in advance. The second preset threshold corresponding to the evaluation metric of the fault prediction model is the performance requirement for the fault prediction model determined based on actual maintenance needs. Furthermore, the effective duration can be adjusted during the fault prediction model training process, within a preset number of iterations, once per preset number of iterations; or it can be adjusted after completing the preset number of iterations, followed by a new round of iterative training based on the adjusted effective duration and model parameters. No specific limitations are imposed on either approach.

[0092] In existing training methods for fault prediction models, model parameters are not flexibly configured during model tuning. For mechanistic and mathematical models, in most cases only the main parameters of the model itself are considered, without paying attention to the parameters that need to be configured during model use. This will cause the trained model to become disconnected from actual implementation and application, requiring the model to be readjusted again during subsequent implementation and deployment.

[0093] In one embodiment, for the initially established fault prediction model, the effective duration corresponding to each type of fault can be set to a large value. Based on this, as long as the model can predict the fault, it is considered an accurate prediction, ensuring a high prediction accuracy. Then, during training, the effective duration value can be gradually decreased, and the model's main parameters can be adjusted to maintain the prediction accuracy at the expected level. At this point, the temporal accuracy may be low, meaning the predicted fault occurrence times are scattered and far from the effective time points; or it may be high, meaning the predicted fault occurrence times are concentrated and tend towards the effective time points. By adjusting the model's main parameters, the temporal accuracy can meet the performance requirements of the fault prediction model. During the training of the fault prediction model, the model's main parameters and configured effective duration, among other parameters, can be adjusted. Parameter tuning during model training is more flexible, and initial model tuning helps the model quickly find optimization directions.

[0094] Furthermore, based on the configuration of the effective duration, the effective prediction results of faults in predictive maintenance can be enhanced. Specifically, the fault prediction model is optimized and pre-trained in actual application. The push time can be different according to the effective duration set for different faults. It can adaptively trigger an alarm based on the predicted time and the predicted fault occurrence time, at the time expected by the user, according to the fault prediction model, to solve the equipment fault problem.

[0095] In this embodiment, based on the effective duration corresponding to different faults, on the one hand, the model evaluation indicators such as prediction accuracy and time accuracy are made more reasonable and accurate, which is conducive to the optimization of model parameters; on the other hand, during the model training process, the main parameters of the model and configuration parameters such as the effective duration can be adjusted, which improves the flexibility of parameter adjustment during the model training process. The parameter optimization not only focuses on the main parameters of the model itself, but also considers the configuration parameters, so that the fault prediction model is more in line with the actual business needs when it is applied in subsequent applications, thereby improving the prediction effect of the fault prediction model, ensuring the timeliness and effectiveness of predictive maintenance, and improving the user experience.

[0096] The predictive maintenance optimization apparatus provided by the present invention will be described below. The predictive maintenance optimization apparatus described below can be referred to in correspondence with the predictive maintenance optimization method described above.

[0097] Reference Figure 3 The predictive maintenance optimization apparatus provided in this embodiment of the invention includes:

[0098] The data acquisition module 10 is used to acquire the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs.

[0099] The fault screening module 20 is used to compare the time difference between the first time point and the second time point with the effective duration corresponding to the target fault, so as to determine the number of accurate predicted faults when the time difference is less than the effective duration.

[0100] The model optimization module 30 is used to calculate the evaluation index of the fault prediction model based on the number of accurately predicted faults, and optimize the fault prediction model according to the evaluation index.

[0101] In one embodiment, the evaluation metric includes time accuracy; the model optimization module 30 is further configured to:

[0102] Based on the target time point of each accurate prediction of the target fault in the number of accurate predicted faults, and the second time point of the actual occurrence of the target fault corresponding to the target time point, the prediction duration of the target fault by the fault prediction model is calculated.

[0103] Calculate the time standard deviation of the fault prediction model based on the predicted duration, the effective duration, and the number of accurately predicted faults.

[0104] The time accuracy of the fault prediction model is obtained by calculating the ratio of the target difference to the effective duration; the target difference is the difference between the effective duration and the time standard deviation.

[0105] In one embodiment, the fault screening module 20 is further configured to:

[0106] The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to obtain the number of effective predicted faults where the predicted time point corresponding to the first time point is before the second time point, and based on the number of effective predicted faults, the number of accurate predicted faults where the time difference is less than the effective duration is determined.

[0107] In one embodiment, the evaluation metric includes prediction accuracy; the model optimization module 30 is further configured to:

[0108] Obtain the union of the number of effective predicted faults and the number of actual occurrences of the target fault;

[0109] The prediction accuracy of the fault prediction model is calculated based on the ratio of the number of accurately predicted faults to the number of faults in the union set.

[0110] In one embodiment, the fault screening module 20 is further configured to:

[0111] Obtain the fault type of the target fault;

[0112] Based on the fault type, determine the effective duration corresponding to the target fault.

[0113] In one embodiment, the effective duration includes device runtime and natural duration.

[0114] In one embodiment, the effective duration varies with gradient; the model optimization module 30 is further configured to:

[0115] During the training process of the fault prediction model, the model parameters of the fault prediction model are adjusted according to the evaluation index to optimize the fault prediction model.

[0116] The effective duration is adjusted based on a preset gradient value. The steps of obtaining the first time point at which the target fault is predicted by the fault prediction model and the second time point at which the target fault actually occurs are returned and executed until the effective duration reaches a first preset threshold and the evaluation index reaches a second preset threshold.

[0117] This invention also provides an engineering machinery, including an engineering machinery body, which is provided with a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the steps of the predictive maintenance optimization method as described in the above embodiments.

[0118] The engineering machinery provided in this embodiment of the invention and the predictive maintenance optimization methods described in the above embodiments can be referred to each other, and will not be repeated here.

[0119] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a predictive maintenance optimization method, which includes:

[0120] Obtain the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs;

[0121] The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to determine the number of accurate predicted faults when the time difference is less than the effective duration.

[0122] Based on the accurate prediction of the number of failures, the evaluation index of the failure prediction model is calculated, and the failure prediction model is optimized according to the evaluation index.

[0123] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0124] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the predictive maintenance optimization method provided by the above methods, the method comprising:

[0125] Obtain the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs;

[0126] The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to determine the number of accurate predicted faults when the time difference is less than the effective duration.

[0127] Based on the accurate prediction of the number of failures, the evaluation index of the failure prediction model is calculated, and the failure prediction model is optimized according to the evaluation index.

[0128] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the aforementioned predictive maintenance optimization methods, the method comprising:

[0129] Obtain the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs;

[0130] The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to determine the number of accurate predicted faults when the time difference is less than the effective duration.

[0131] Based on the accurate prediction of the number of failures, the evaluation index of the failure prediction model is calculated, and the failure prediction model is optimized according to the evaluation index.

[0132] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0133] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0134] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A predictive maintenance optimization method, characterized in that, include: Obtain the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs; Obtain the fault type of the target fault; Based on the fault type, determine the effective duration corresponding to the target fault; The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to determine the number of accurate predicted faults when the time difference is less than the effective duration. Based on the accurate prediction of the number of failures, the evaluation index of the failure prediction model is calculated, and the failure prediction model is optimized according to the evaluation index.

2. The predictive maintenance optimization method according to claim 1, characterized in that, The evaluation metrics include time accuracy; the evaluation metrics for the fault prediction model, calculated based on the accurate prediction of the number of faults, include: Based on the target time point of each accurate prediction of the target fault in the number of accurate predicted faults, and the second time point of the actual occurrence of the target fault corresponding to the target time point, the prediction duration of the target fault by the fault prediction model is calculated. Calculate the time standard deviation of the fault prediction model based on the predicted duration, the effective duration, and the number of accurately predicted faults. The time accuracy of the fault prediction model is obtained by calculating the ratio of the target difference to the effective duration; the target difference is the difference between the effective duration and the time standard deviation.

3. The predictive maintenance optimization method according to claim 1, characterized in that, The step of comparing the time difference between the first time point and the second time point with the effective duration corresponding to the target fault to determine the accurate number of predicted faults where the time difference is less than the effective duration includes: The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to obtain the number of effective predicted faults where the predicted time point corresponding to the first time point is before the second time point, and based on the number of effective predicted faults, the number of accurate predicted faults where the time difference is less than the effective duration is determined.

4. The predictive maintenance optimization method according to claim 3, characterized in that, The evaluation metrics include prediction accuracy; the evaluation metrics for the fault prediction model, calculated based on the number of accurately predicted faults, include: Obtain the union of the number of effective predicted faults and the number of actual occurrences of the target fault; The prediction accuracy of the fault prediction model is calculated based on the ratio of the number of accurately predicted faults to the number of faults in the union set.

5. The predictive maintenance optimization method according to claim 4, characterized in that, The effective duration includes both device runtime and natural duration.

6. The predictive maintenance optimization method according to claim 1, characterized in that, The effective duration varies with a gradient, and the optimization of the fault prediction model based on the evaluation index includes: During the training process of the fault prediction model, the model parameters of the fault prediction model are adjusted according to the evaluation index to optimize the fault prediction model. The effective duration is adjusted based on a preset gradient value. The steps of obtaining the first time point at which the target fault is predicted by the fault prediction model and the second time point at which the target fault actually occurs are returned and executed until the effective duration reaches a first preset threshold and the evaluation index reaches a second preset threshold.

7. A predictive maintenance optimization device, characterized in that, include: The data acquisition module is used to acquire the first time point at which the target fault is predicted to occur by the fault prediction model, and the second time point at which the target fault actually occurs. The fault screening module is used to obtain the fault type of the target fault; Based on the fault type, determine the effective duration corresponding to the target fault; The time difference between the first time point and the second time point is compared with the effective duration corresponding to the target fault to determine the number of accurate predicted faults when the time difference is less than the effective duration. The model optimization module is used to calculate the evaluation index of the fault prediction model based on the number of accurately predicted faults, and to optimize the fault prediction model according to the evaluation index.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the predictive maintenance optimization method as described in any one of claims 1 to 6.

9. An engineering machinery, characterized in that, The invention includes a construction machinery body, which is equipped with a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the steps of the predictive maintenance optimization method as described in any one of claims 1 to 6.

Images