A method for predicting bearing faults in rotating mechanical equipment

Abstract

The present invention relates to a method for predicting bearing faults of rotating mechanical equipment, belonging to the technical field of bearing fault prediction. The method includes: interacting with big data of bearing faults based on the scene feature information of the target scene to obtain a fault record set; performing clustering division on the fault record set based on fault categories, and analyzing the clustering division results to obtain a multi-condition prediction reference index set; according to the multi-condition prediction reference index set, performing prediction resource allocation for multiple conditions to obtain a prediction inclination coefficient set for multiple conditions; combining the fault record set with the prediction inclination coefficient set to construct and train a bearing fault prediction model, wherein the bearing fault prediction model includes multiple prediction sub-networks, and the multiple prediction sub-networks have condition marks; through the bearing fault prediction model, performing fault prediction based on the real-time monitoring data of the bearing to obtain a fault prediction result. Furthermore, the technical effects of online monitoring, complete monitoring, and good early warning ability are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of bearing fault detection, and particularly to a method for predicting bearing faults in rotating mechanical equipment. Background Art

[0002] Rotating mechanical equipment is widely used in various industries, including manufacturing, transportation, aviation, energy and other fields. As a key component of rotating machinery, the operating state of the bearing directly affects the safety, reliability and operating efficiency of the equipment. Accurately and timely predicting bearing faults, especially detecting potential problems early during equipment operation, is an important task for improving equipment reliability, reducing maintenance costs and ensuring production safety.

[0003] Currently, common bearing fault prediction methods mostly rely on traditional signal processing techniques and basic model analysis methods, such as vibration analysis, temperature monitoring and noise monitoring. These methods analyze the physical signals of the equipment to determine whether there are faults. There are technical problems such as relying on expert experience, lacking adaptability, and having poor performance when dealing with complex problems of multiple working conditions and multiple fault types. Summary of the Invention

[0004] In view of the technical problems in the prior art of relying on expert experience, lacking adaptability, and having poor performance when dealing with complex problems of multiple working conditions and multiple fault types, the present invention provides a method for predicting bearing faults in rotating mechanical equipment to solve these problems.

[0005] The technical solution of the present invention to solve the above technical problems is as follows:

[0006] In a first aspect, the present invention provides a method for predicting bearing faults in rotating mechanical equipment.

[0007] Based on the scenario feature information of the target scenario, interact with big data of bearing faults to obtain a fault record set.

[0008] Perform clustering division on the fault record set based on fault categories, and perform frequency dimension analysis, intensity dimension analysis and breadth dimension analysis based on the clustering division results to obtain a multi-condition prediction reference index set.

[0009] According to the multi-condition prediction reference index set, perform prediction resource allocation for multiple working conditions to obtain a set of prediction inclination coefficients for multiple working conditions, where the set of prediction inclination coefficients includes sample inclination coefficients, call inclination coefficients and scale inclination coefficients for multiple working conditions.

[0010] Combine the fault record set and the set of prediction inclination coefficients to construct and train a bearing fault prediction model, where the bearing fault prediction model includes multiple prediction sub-networks, and multiple of the prediction sub-networks have working condition marks.

[0011] Through the bearing fault prediction model, perform fault prediction based on real-time monitoring data of the bearing to obtain a fault prediction result.

[0012] In a feasible implementation manner, the fault record set includes an original fault record set of historical fault records based on a target scenario and a homologous fault record set of historical fault records of homologous scenarios based on the target scenario.

[0013] In a feasible implementation manner, perform clustering division on the fault record set based on fault categories, and perform frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis based on the clustering division result to obtain a multi-condition prediction reference index set, including:

[0014] Taking multiple fault categories as division targets, perform clustering division on the fault record set to generate multiple working condition fault record sets.

[0015] Traverse multiple working condition fault record sets, count the frequency of fault records for frequency dimension analysis, and obtain a frequency reference index set.

[0016] Traverse multiple working condition fault record sets, obtain the typical prediction accuracy rates of multiple fault categories for intensity dimension, and obtain an intensity reference index set.

[0017] Traverse multiple working condition fault record sets, evaluate the fault influence of multiple fault categories for breadth dimension analysis, and obtain a breadth reference index set.

[0018] Output the frequency reference index set, the intensity reference index set, and the breadth reference index set as the multi-condition prediction reference index set.

[0019] In a feasible implementation manner, according to the multi-condition prediction reference index set, perform prediction resource allocation for multiple working conditions to obtain a prediction inclination coefficient set for multiple working conditions, where the prediction inclination coefficient set includes sample inclination coefficients, call inclination coefficients, and scale inclination coefficients for multiple working conditions, including:

[0020] According to the frequency reference index set and the breadth reference index set, determine the sample inclination coefficients of multiple working conditions.

[0021] According to the frequency reference index set and the intensity reference index set, determine the scale inclination coefficients of multiple working conditions.

[0022] According to the intensity reference index set and the breadth reference index set, determine the call inclination coefficients of multiple working conditions.

[0023] In a feasible implementation manner, combine the fault record set and the prediction inclination coefficient set to construct and train a bearing fault prediction model, including:

[0024] Construct multiple discriminant networks based on machine learning, where the multiple discriminant networks are constructed based on different machine learning models.

[0025] Based on the call tilt coefficient, define the proportion of the number of networks corresponding to multiple working conditions, and divide the multiple discriminant networks into multiple groups of working condition discriminant networks.

[0026] Based on the scale tilt coefficient, adjust the network depth of the multiple groups of working condition discriminant networks.

[0027] Based on the sample tilt coefficient, define the sample capacity of the multiple groups of working condition discriminant networks, and extract multiple working condition sample sets corresponding to the fault record set.

[0028] Through the multiple working condition sample sets, perform supervised training on the multiple groups of working condition discriminant networks after network depth adjustment, obtain multiple prediction sub-networks with working condition labels, and generate the bearing fault prediction model through an ensemble learning method.

[0029] In a feasible implementation manner, through the bearing fault prediction model, perform fault prediction based on real-time bearing monitoring data to obtain a fault prediction result. After that, the method further includes:

[0030] Record the fault prediction result and construct a prediction record library.

[0031] According to the prediction record library, configure an accumulator to accumulate false fault prediction results of multiple working conditions, where the false fault prediction results include false alarm fault prediction results and missed alarm fault prediction results.

[0032] When the accumulated result of the accumulator is greater than or equal to a preset false alarm adjustment threshold, call the multiple false fault prediction results corresponding to the accumulated result as a false fault record set to optimize the bearing fault prediction model.

[0033] In a feasible implementation manner, calling the multiple false fault prediction results corresponding to the accumulated result as a false fault record set to optimize the bearing fault prediction model includes:

[0034] Obtain the fault handling information of the multiple false fault prediction results.

[0035] According to the fault handling information, perform prediction correction on the multiple false fault prediction results to obtain an improved fault record set.

[0036] Use the improved fault record set to perform enhanced supervised training on the bearing fault prediction model.

[0037] In a feasible implementation manner, strengthening the supervised training of the bearing fault prediction model with the improved fault record set includes:

[0038] Based on the false fault record set, extract the first false fault working conditions, and based on the improved fault record set, extract the corresponding first improved fault record subset.

[0039] Call multiple first working condition prediction sub-networks corresponding to the first false fault working conditions, and perform model mutation and parameter exchange based on the multiple first working condition prediction sub-networks to obtain N first supplementary prediction sub-networks, where N is a positive integer greater than or equal to 2.

[0040] Use the first improved fault record subset as training data to perform enhanced training on the multiple first working condition prediction sub-networks and the N first supplementary prediction sub-networks, obtain multiple first working condition enhanced prediction sub-networks and evaluate the prediction performance.

[0041] According to the prediction performance evaluation results, exclude the last N first working condition enhanced prediction sub-networks with the worst prediction performance, and synchronously retain the multiple first working condition enhanced prediction sub-networks to the bearing fault prediction model.

[0042] In a second aspect, the present invention also provides an electronic device, including: a memory for storing executable instructions; a processor for implementing a method for predicting bearing faults of a rotating mechanical device provided by the present invention when executing the executable instructions stored in the memory.

[0043] In a third aspect, the present invention also provides a computer-readable storage medium storing a computer program, which when executed by a processor, implements a method for predicting bearing faults of a rotating mechanical device provided by the present invention.

[0044] The beneficial effects of the present invention are as follows: Through the scene feature information of the target scene, combined with the big data of interactive bearing faults, a fault record set is obtained. Perform clustering analysis on the fault record set based on fault categories, and perform analysis in the dimensions of frequency, intensity, and breadth according to the clustering results to obtain a multi-working condition prediction reference index set. According to the multi-working condition prediction reference index set, perform prediction resource allocation for multiple working conditions to obtain a prediction inclination coefficient set, and the inclination coefficient set includes a sample inclination coefficient, a call inclination coefficient, and a scale inclination coefficient. Combine the fault record set with the prediction inclination coefficient set to construct and train a bearing fault prediction model, which includes multiple prediction sub-networks and marks the working conditions for each sub-network. Through the bearing fault prediction model, perform fault prediction based on the real-time monitoring data of the bearing to generate a fault prediction result. The technical effects of improving the self-adaptability of bearing fault prediction and improving the performance when dealing with multiple working conditions and multiple fault types are achieved. Description of the Drawings

[0045] Figure 1 Schematic flow diagram of a bearing fault prediction method for rotating mechanical equipment provided by the present invention;

[0046] Figure 2 Schematic flow diagram of constructing and training a bearing fault prediction model in a bearing fault prediction method for rotating mechanical equipment provided by the present invention;

[0047] Figure 3 Schematic structural diagram of an electronic device provided by the present invention;

[0048] Figure 4 Schematic structural diagram of a computer-readable storage medium provided by the present invention.

[0049] In the drawings, the list of components represented by each reference numeral is as follows:

[0050] Electronic device 500, memory 510, processor 520, first computer program 511, computer-readable storage medium 600, second computer program 611. Detailed implementation manners

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present invention.

[0052] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0053] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein. Embodiment 1

[0054] As Figure 1 shown, an embodiment of the present invention provides a method for predicting bearing faults in rotating mechanical equipment.

[0055] Based on the scenario feature information of the target scenario, interact with the big data of bearing faults to obtain a fault record set.

[0056] Specifically, the target scenario refers to the operating state under the actual environment or conditions that need to be analyzed and processed in a specific application. In bearing fault prediction, the target scenario is a specific bearing and its corresponding operating environment. Among them, the scenario feature information refers to the data describing the relevant factors in the target scenario. Exemplarily, it includes: environmental data such as temperature, humidity, and air pressure; mechanical parameters such as rotational speed, load, and vibration frequency; operation parameters such as working state and operation cycle.

[0057] Specifically, the big data of bearing faults includes a historical data set of bearing faults under multiple interaction factors (such as mechanical operating state, environmental conditions, operation conditions, etc.). In other words, this big data of bearing faults includes bearing operation data from different time points and different working conditions, and contains rich fault modes and fault signals.

[0058] Specifically, the fault record set is a set composed of historical fault data. This fault record set is extracted from the above-mentioned big data of bearing faults and contains the complex mapping relationship between the bearing operation state data and the fault prediction result, which is used to train, test, and evaluate the fault diagnosis model, and further predict future fault events.

[0059] In some embodiments, the fault record set includes an original fault record set of historical fault records based on the target scenario and a homologous fault record set of historical fault records of homologous scenarios based on the target scenario.

[0060] Furthermore, the fault record set includes an original fault record set and a homologous fault record set. Among them, the original fault record set refers to the historical fault records based on the target scenario, including the fault data that occurred in the target scenario. For example, the record data of the faults that occurred in bearings of the same model under the same temperature and load conditions; the homologous fault record set refers to the historical fault records of the homologous scenarios based on the target scenario. The homologous scenario is a scenario with a similar working environment or operating conditions to the target scenario, that is, the fault modes in the target scenario may have a high similarity with the fault modes in these homologous scenarios. Exemplarily, the data with similar operating conditions and fault modes are classified by data clustering methods (such as K-means, DBSCAN, etc.), and the fault records of the similar scenarios are extracted as the homologous fault record set.

[0061] By combining the original fault records and the homologous fault records, the diversity of the training set can be expanded, the overfitting phenomenon can be reduced, the adaptability of the model in different operating environments can be improved, and at the same time, it helps to better understand the similarities of different fault modes and causes, thereby improving the accuracy of fault diagnosis and prediction.

[0062] Perform clustering division on the fault record set based on fault categories, and perform frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis based on the clustering division results to obtain a multi-condition prediction reference index set.

[0063] Specifically, the fault events in the fault record set are divided into different categories according to similarity, so that each category of fault events has high similarity in characteristics. Different types of fault modes, characteristics, and rules can be identified through clustering, providing a basis for subsequent analysis and prediction.

[0064] Specifically, through frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis, the comprehensive multi-condition prediction reference index set obtained helps to better understand the characteristics and impacts of different categories of faults. This multi-condition prediction reference index set is used to guide the construction of the prediction model subsequently, so that the prediction model adapts to the fault characteristics of the target scenario and improves the fault prediction accuracy and response ability under different working conditions.

[0065] In some embodiments, performing clustering division on the fault record set based on fault categories, and performing frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis based on the clustering division results to obtain a multi-condition prediction reference index set includes:

[0066] Taking multiple fault categories as the division target, clustering and dividing the fault record set to generate multiple working condition fault record sets; traversing multiple working condition fault record sets, counting the fault record frequencies for frequency dimension analysis to obtain a frequency reference index set; traversing multiple working condition fault record sets, obtaining the typical prediction accuracies of multiple fault categories for intensity dimension to obtain an intensity reference index set; traversing multiple working condition fault record sets, evaluating the fault impact of multiple fault categories for breadth dimension analysis to obtain a breadth reference index set; outputting the frequency reference index set, the intensity reference index set and the breadth reference index set as the multi-working condition prediction reference index set.

[0067] Specifically, first, by clustering the fault record set, faults of similar types are classified into the same fault prediction working condition to generate multiple working condition fault record sets. In other words, each working condition fault record set represents the fault record data under a certain fault category.

[0068] Specifically, the frequency dimension analysis focuses on the occurrence probability or frequency of different fault categories, so as to analyze the occurrence frequency of each category of faults and generate a frequency reference index set, which characterizes the universality and commonness of fault categories. Exemplarily, for each working condition fault record set, statistical methods such as frequency statistics and proportion statistics are used to calculate the occurrence times and occurrence frequencies of each fault category in the set as the intensity reference index of multiple working condition fault record sets.

[0069] Specifically, the intensity dimension analysis focuses on the prediction difficulty of different fault categories. In other words, the intensity reference index set is a quantitative representation of the prediction difficulty of different fault categories. Fault categories with greater prediction difficulty often involve more complex fault mechanisms and require deeper neural network structures for modeling. Among them, the prediction difficulty of fault prediction is mainly closely related to the complexity of fault categories and the characteristics of data. Therefore, the prediction difficulty can be quantified based on the feature complexity of faults. For example, some fault categories may contain multiple sub-fault modes and have a higher prediction difficulty. Through intensity dimension analysis, an appropriate network depth and structure can be selected for the prediction tasks of different categories of faults.

[0070] Specifically, the breadth - dimension analysis focuses on the impact of different fault categories. The impact involves the scope and severity of the fault on the system or equipment. Fault categories with high severity and wide impact scope have greater impact and need more attention in the prediction model. Exemplarily, first, through the fault labels in the historical fault record set, analyze the impact scope of the fault (e.g., whether it causes equipment downtime, whether it affects other components) and severity (e.g., whether the fault causes equipment damage, repair cost, etc.); then, use methods such as weighted average method and comprehensive scoring method, combine data from different dimensions, calculate the equivalent impact intensity of each fault category, and generate a comprehensive score of the impact intensity as the breadth reference index set.

[0071] Optionally, through the breadth - dimension analysis, priorities can be set for different fault categories. For example, fault categories with large impact scope and high severity (such as faults that cause complete equipment damage) should be allocated more prediction resources preferentially to reduce potential economic losses.

[0072] Generally speaking, through the multi - level analysis of the three dimensions of frequency, intensity, and breadth, more accurate analysis and prediction can be carried out for different fault categories in the fault record set. This not only helps to better understand the characteristics and impacts of different types of faults, but also provides clear guidance for model design and resource allocation, thereby improving the fault prediction accuracy and response ability.

[0073] According to the multi - condition prediction reference index set, allocate prediction resources for multiple conditions to obtain a set of prediction skew coefficients for multiple conditions, where the set of prediction skew coefficients includes sample skew coefficients, call skew coefficients, and scale skew coefficients for multiple conditions.

[0074] Specifically, based on the frequency reference index set, intensity reference index set, and breadth reference index set in the multi - condition prediction reference index set, define the skew levels of prediction resources for different conditions, namely sample skew coefficients, call skew coefficients, and scale skew coefficients.

[0075] Specifically, the sample skew coefficient quantifies the skew degree of the sample distribution of each condition, which helps to make adjustments for the sample imbalance problem and avoid the model biasing towards majority - class samples; the call skew coefficient quantifies the model attention corresponding to each condition, ensuring that tasks with high impact and complexity have high model attention. In other words, for conditions with higher call skew coefficients, preferentially allocate more computing resources or more efficient model call strategies (such as a larger number of prediction sub - networks); the scale skew coefficient defines the depth of the prediction model for each condition. In other words, the larger the scale skew coefficient, the greater the prediction difficulty of the model, and a more complex prediction model is required to better express the occurrence characteristics of faults, thereby improving the prediction accuracy.

[0076] In some embodiments, according to the multi-condition prediction reference index set, perform prediction resource allocation for multiple conditions to obtain a prediction inclination coefficient set for multiple conditions, where the prediction inclination coefficient set includes sample inclination coefficients, call inclination coefficients, and scale inclination coefficients for multiple conditions, including:

[0077] Determine the sample inclination coefficients for multiple conditions according to the frequency reference index set and the breadth reference index set; determine the scale inclination coefficients for multiple conditions according to the frequency reference index set and the intensity reference index set; determine the call inclination coefficients for multiple conditions according to the intensity reference index set and the breadth reference index set.

[0078] Specifically, the sample inclination coefficient is evaluated and obtained based on the relationship between the fault occurrence frequency and the fault influence range. A condition with a higher frequency reference index means that the fault of this category has a higher occurrence frequency, and undersampling needs to be performed to appropriately reduce the data volume of the sample data, balance the imbalance between classes, and prevent the model from biasing towards the majority class. A higher breadth reference index represents a greater influence of this fault category, and it may be necessary to correspondingly improve the data quality of the sample data.

[0079] Specifically, the scale inclination coefficient mainly reflects the prediction difficulty of the fault type and its requirement for the model depth. By combining the intensity reference index (i.e., difficulty and complexity) and the frequency reference index (i.e., occurrence frequency) of the fault, it is determined how deep a model is required for this condition to perform effective prediction. Exemplarily, if the fault type of certain conditions occurs frequently but has a high prediction difficulty, a deeper model (such as a deep neural network, LSTM, CNN, etc.) may be required to process it to improve the fault prediction accuracy; if a certain condition has a high frequency but a low prediction difficulty, a too deep model structure does not need to be configured for this condition to reduce the demand for computing resources.

[0080] Specifically, the call inclination coefficient reflects the prediction attention of the fault category, that is, which conditions require more computing resources or a more efficient call strategy in a multi-task environment. Exemplarily, if the fault type of certain conditions has a low occurrence frequency but a high prediction difficulty, while using a model network with a large depth for it, reduce the number of prediction sub-networks facing this condition in the ensemble learning model to avoid over-occupying prediction resources.

[0081] By constructing a prediction inclination coefficient set through the combination of the frequency reference index set, the intensity reference index set, and the breadth reference index set, it can help achieve optimal resource allocation, dynamically adjust the resource configuration of the model according to the characteristics of different conditions, and thus improve the overall prediction accuracy and efficiency.

[0082] Construct and train a bearing fault prediction model by combining the fault record set with the predicted tilt coefficient set, where the bearing fault prediction model includes multiple prediction sub-networks, and multiple said prediction sub-networks have operating condition labels.

[0083] Optionally, based on the ensemble learning method, construct and train a bearing fault prediction model including multiple prediction sub-networks by combining the fault record set with the predicted tilt coefficient set. The multiple prediction sub-networks correspond to different operating condition categories, and this corresponding relationship is characterized by operating condition labels.

[0084] In some embodiments, as Figure 2 shown, constructing and training a bearing fault prediction model by combining the fault record set with the predicted tilt coefficient set includes:

[0085] Construct multiple discriminant networks based on machine learning, where multiple said discriminant networks are constructed based on different machine learning models; define the proportion of the number of networks corresponding to multiple operating conditions based on the call tilt coefficient, and divide multiple said discriminant networks into multiple groups of operating condition discriminant networks; adjust the network depth of multiple said groups of operating condition discriminant networks based on the scale tilt coefficient; define the sample size of multiple said groups of operating condition discriminant networks based on the sample tilt coefficient, and extract multiple operating condition sample sets corresponding to the fault record set; perform supervised training on multiple said groups of operating condition discriminant networks with adjusted network depth through multiple operating condition sample sets, obtain multiple said prediction sub-networks with operating condition labels, and generate the bearing fault prediction model through the ensemble learning method.

[0086] Specifically, first, construct multiple discriminant networks based on machine learning for fault prediction. The structures of multiple discriminant networks can be different machine learning models, such as support vector machines (SVMs), decision trees, deep neural networks, etc., to adapt to different operating condition requirements.

[0087] Specifically, based on the call tilt coefficient, define the proportion of the number of networks corresponding to different operating conditions, reasonably allocate the computing resources and task complexity of the model, optimize the resource usage of the model, and then generate multiple corresponding groups of operating condition discriminant networks. Defining the proportion of the number of networks corresponding to each operating condition is to allocate a corresponding number of discriminant networks to each operating condition. For example, an operating condition with a higher call tilt coefficient may be allocated multiple networks to ensure that it receives sufficient attention during training and prediction.

[0088] Specifically, according to the scale inclination coefficient of each working condition, the discriminant network groups for different working conditions are deeply adjusted to ensure that more complex fault modes have sufficient network levels and depths for effective prediction. Exemplarily, for working conditions with a higher scale inclination coefficient, increase the number of layers (i.e., network depth) and nodes of the discriminant network, or switch to a more complex model, such as a deep convolutional neural network or a long short-term memory network. For working conditions with a lower scale inclination coefficient, a shallow network or other lightweight models can be used to reduce the consumption of computing resources.

[0089] Specifically, sample data is allocated to different working conditions according to the sample inclination coefficient to avoid the problem of sample imbalance, thereby improving the generalization ability of the model. Exemplarily, for fault types (working conditions) with fewer samples (relatively less likely to occur) but having an important impact, oversampling and other techniques are used to increase the diversity of training data to ensure that the network can obtain sufficient sample support.

[0090] Furthermore, supervised training of the discriminant network groups for working conditions after network depth adjustment is performed according to different working condition sample sets. Each discriminant network group for working conditions learns the fault modes of specific working conditions based on its corresponding working condition sample set, and multiple prediction sub-networks with working condition labels are obtained. Then, these prediction sub-networks are combined through ensemble learning (such as weighted voting, stacking methods, etc.) to generate a bearing fault prediction model, which can perform fault prediction under different working conditions and provide reliable fault diagnosis support.

[0091] Through the bearing fault prediction model, fault prediction is performed based on the real-time monitoring data of the bearing to obtain the fault prediction result.

[0092] Specifically, first, the real-time monitoring data of the bearing is obtained. The real-time monitoring data is collected through sensors or monitoring systems and includes various parameters related to the running state of the bearing, such as temperature, vibration, pressure, rotational speed, noise, etc. Then, the collected real-time data is preprocessed, and key features helpful for fault prediction are extracted. Next, the extracted features are input into the bearing fault prediction model for fault prediction to obtain the fault prediction result. Optionally, the fault prediction result includes the fault probability, fault level, fault type, etc.

[0093] Furthermore, through the bearing fault prediction model, fault prediction is performed based on the real-time monitoring data of the bearing to obtain the fault prediction result. After that, the method further includes:

[0094] Record the fault prediction results and construct a prediction record library; according to the prediction record library, configure an accumulator to accumulate the false fault prediction results under multiple working conditions, where the false fault prediction results include false alarm fault prediction results and missed alarm fault prediction results; when the accumulated result of the accumulator is greater than or equal to a preset false alarm adjustment threshold, call the multiple false fault prediction results corresponding to the accumulated result as a false fault record set to optimize the bearing fault prediction model.

[0095] Specifically, after obtaining the fault prediction results, store the fault prediction results in the constructed prediction record library, which includes multiple historical fault prediction results; then, configure an accumulator to accumulate the false fault prediction results under different working conditions to detect the error of the prediction model and prepare for subsequent optimization.

[0096] Specifically, the false fault record set contains false alarm fault prediction results (false positives) and missed alarm fault prediction results (false negatives), such as false alarms (the model incorrectly predicts the occurrence of a fault, but in fact the bearing has not failed) and missed alarms (the model fails to predict an actual fault).

[0097] Specifically, set a false alarm adjustment threshold. If the accumulated false fault prediction results of a certain working condition reach this threshold, it means that there are significant false alarm or missed alarm problems in the prediction results of this working condition and need to be optimized. Take the corresponding false fault prediction results (including false alarm and missed alarm prediction results) as the false fault record set.

[0098] Specifically, retrain or fine-tune the prediction model through the false fault record set to reduce the probability of false alarms and missed alarms and enhance the generalization ability and robustness of the model. Optionally, the implementation methods for optimizing the bearing fault prediction model include incremental learning, model architecture adjustment, loss function improvement, etc.

[0099] In some embodiments, calling the multiple false fault prediction results corresponding to the accumulated result as a false fault record set to optimize the bearing fault prediction model includes:

[0100] Obtain the fault handling information of the multiple false fault prediction results; perform prediction correction on the multiple false fault prediction results according to the fault handling information to obtain an improved fault record set; use the improved fault record set to perform enhanced supervised training on the bearing fault prediction model.

[0101] Specifically, first, obtain the fault handling information corresponding to multiple false fault prediction results. This fault handling information is a record of the response and handling of false fault prediction results, including eliminating false alarms, supplementing missed alarms, etc.; then, correct the false fault prediction results according to the fault handling information. This includes adjusting the predicted fault type, fault degree, fault time, etc. The corrected false fault prediction results are the improved fault record set, which can be considered as the correct fault prediction record data; furthermore, use the improved fault record set to perform enhanced supervised training on the bearing fault prediction model, including retraining the model with new training data, or optimizing the model using reinforcement learning algorithms, and adjusting the model parameters to improve the accuracy of fault prediction.

[0102] In some implementation manners, performing enhanced supervised training on the bearing fault prediction model with the improved fault record set includes:

[0103] Based on the false fault record set, extract the first false fault working conditions, and based on the improved fault record set, extract the corresponding first improved fault record subset; call multiple first working condition prediction subnets corresponding to the first false fault working conditions, and perform model mutation and parameter exchange based on the multiple first working condition prediction subnets to obtain N first supplementary prediction subnets, where N is a positive integer greater than or equal to 2; use the first improved fault record subset as training data to perform enhanced training on the multiple first working condition prediction subnets and the N first supplementary prediction subnets to obtain multiple first working condition enhanced prediction subnets and evaluate the prediction performance; according to the prediction performance evaluation results, exclude the last N first working condition enhanced prediction subnets with the worst prediction performance, and synchronously retain the multiple first working condition enhanced prediction subnets to the bearing fault prediction model.

[0104] Specifically, first randomly extract the first false fault working conditions from the false fault record set, and at the same time extract the corresponding first improved fault record subset from the improved fault record set. The first improved fault record set is a data set optimized or corrected from the false fault record set, that is, additional annotations or corrections are made on the basis of the false fault record set to more realistically reflect the fault mode under the working conditions.

[0105] Optionally, use statistical analysis methods, such as calculating indicators such as false alarm rate and missed alarm rate, and select the working conditions with higher false alarm or missed alarm probabilities as the first false fault working conditions to ensure that the working conditions with poor model prediction effects are improved first.

[0106] Specifically, the multiple first working condition prediction sub-networks are multiple initial networks that have been trained for the first false fault working condition. By performing network mutation (such as modifying the network structure, changing the activation function, etc.) and parameter exchange (such as sharing weights between different network layers, changing the initialization method, etc.) on the multiple first working condition prediction sub-networks, N first supplementary prediction sub-networks are generated. The purpose of the N first supplementary prediction sub-networks is to increase the diversity of the network and prevent the model from falling into a local optimal solution.

[0107] Specifically, use the first improved fault record subset as training data to perform reinforcement training on the multiple first working condition prediction sub-networks and the first supplementary prediction sub-networks, that is, adjust the weights of the network according to the reinforcement learning method to make it better cope with the false fault working condition and improve the prediction ability of the multiple first working condition prediction sub-networks and the first supplementary prediction sub-networks for the false fault working condition.

[0108] Furthermore, after the reinforcement training, through the validation set or cross-validation, evaluate the prediction performance of each first working condition reinforced prediction sub-network. Exemplarily, evaluate the performance of each network according to indicators such as accuracy, recall rate, F1 score, etc.; then, exclude the N first working condition reinforced prediction sub-networks with the worst performance, retain the first working condition reinforced prediction sub-network with the best performance, and integrate it into the bearing fault prediction model.

[0109] Optionally, based on the above-mentioned reinforcement supervised training and ensemble learning methods, traverse the false fault record set to optimize the model for multiple working conditions, so as to comprehensively improve the adaptability of the model to the false fault working condition.

[0110] The reinforcement supervised training process of the above-mentioned bearing fault prediction model is based on the false fault record set and uses the improved fault record set to strengthen the training effect of the model. This process combines multiple network mutations, parameter exchanges, reinforcement training, and performance evaluation to optimize the model, and finally improves the accuracy and stability of fault prediction.

[0111] Through the above method steps, the accuracy and robustness of the bearing fault prediction model in the face of false fault working conditions can be effectively improved, thus ensuring efficient fault prediction in a complex environment.

[0112] A bearing fault prediction method provided by an embodiment of the present invention has at least the following technical effects:

[0113] Based on the scene feature information of the target scene and combined with the big data of interactive bearing faults, a fault record set is obtained. Cluster analysis is performed on the fault record set based on fault categories, and frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis are carried out according to the clustering results to obtain a multi-condition prediction reference index set. According to the multi-condition prediction reference index set, prediction resource allocation for multiple conditions is performed to obtain a prediction inclination coefficient set, where the inclination coefficient set includes a sample inclination coefficient, a call inclination coefficient, and a scale inclination coefficient for each condition. Combining the fault record set and the prediction inclination coefficient set, a bearing fault prediction model is constructed and trained. The model includes multiple prediction sub-networks, and each sub-network is marked with a working condition. Through the bearing fault prediction model, fault prediction is performed based on the real-time monitoring data of the bearing to generate a fault prediction result. Thus, the technical effect of improving the adaptability of bearing fault prediction and improving the performance when dealing with multiple working conditions and multiple fault types is achieved. Embodiment 2

[0114] Please refer to Figure 3 , Figure 3 which is a schematic diagram of an embodiment of an electronic device provided by an embodiment of the present invention. As Figure 3 shown, an embodiment of the present invention provides an electronic device 500, including a memory 510, a processor 520, and a first computer program 511 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the first computer program 511, the following steps are implemented: Based on the scene feature information of the target scene and the big data of interactive bearing faults, a fault record set is obtained; clustering division based on fault categories is performed on the fault record set, and frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis are performed based on the clustering division results to obtain a multi-condition prediction reference index set; according to the multi-condition prediction reference index set, prediction resource allocation for multiple conditions is performed to obtain a prediction inclination coefficient set for multiple conditions, where the prediction inclination coefficient set includes a sample inclination coefficient, a call inclination coefficient, and a scale inclination coefficient for each of the multiple conditions; combining the fault record set and the prediction inclination coefficient set, a bearing fault prediction model is constructed and trained, where the bearing fault prediction model includes multiple prediction sub-networks, and multiple of the prediction sub-networks have working condition markings; through the bearing fault prediction model, fault prediction is performed based on the real-time monitoring data of the bearing to obtain a fault prediction result. Embodiment 3

[0115] Please refer to Figure 4 , Figure 4 which is a schematic diagram of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. As Figure 4As shown in the figure, this embodiment provides a computer-readable storage medium 600, on which a second computer program 611 is stored. When the second computer program 611 is executed by a processor, the following steps are implemented: Based on the scenario feature information of the target scenario, interact with the big data of bearing faults to obtain a fault record set; perform clustering division on the fault record set based on fault categories, and perform frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis based on the clustering division results to obtain a multi-condition prediction reference index set; according to the multi-condition prediction reference index set, perform prediction resource allocation for multiple conditions to obtain a prediction inclination coefficient set for multiple conditions, where the prediction inclination coefficient set includes sample inclination coefficients, call inclination coefficients, and scale inclination coefficients for multiple conditions; combine the fault record set and the prediction inclination coefficient set to construct and train a bearing fault prediction model, where the bearing fault prediction model includes multiple prediction sub-networks, and multiple of the prediction sub-networks have condition marks; through the bearing fault prediction model, perform fault prediction based on the real-time monitoring data of the bearing to obtain a fault prediction result.

[0116] It should be noted that in the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0117] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0119] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one or more of the processes Figure 1 and / or boxes Figure 1 specified in one or more of the processes or boxes

[0120] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes Figure 1 and / or boxes Figure 1 specified in one or more of the boxes or boxes

[0121] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept

[0122] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention is also intended to include these modifications and variations

Claims

1. A method for predicting bearing faults in rotating mechanical equipment, characterized in that, The method includes: Interacting with big data on bearing faults based on the scene feature information of the target scene to obtain a fault record set; Performing clustering division on the fault record set based on fault categories, and performing frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis based on the clustering division results to obtain a multi-condition prediction reference index set; According to the multi-condition prediction reference index set, perform prediction resource allocation for multiple conditions to obtain a prediction inclination coefficient set for multiple conditions, where the prediction inclination coefficient set includes sample inclination coefficients, call inclination coefficients, and scale inclination coefficients for multiple conditions; Combining the fault record set and the prediction inclination coefficient set to construct and train a bearing fault prediction model, where the bearing fault prediction model includes multiple prediction sub-networks, and multiple of the prediction sub-networks have condition markers; Through the bearing fault prediction model, perform fault prediction based on the real-time monitoring data of the bearing to obtain a fault prediction result; Among them, combining the fault record set and the prediction inclination coefficient set to construct and train a bearing fault prediction model includes: Constructing multiple discriminant networks based on machine learning, where multiple of the discriminant networks are constructed based on different machine learning models and at least include support vector machines, decision trees, and deep neural networks; Based on the call inclination coefficient, defining the proportion of the number of networks corresponding to multiple conditions, and dividing the multiple discriminant networks into multiple condition discriminant network groups; Based on the scale inclination coefficient, perform network depth adjustment on the multiple condition discriminant network groups; Based on the sample inclination coefficient, define the sample capacity of the multiple condition discriminant network groups, and extract multiple condition sample sets corresponding to the fault record set; Through multiple condition sample sets, perform supervised training on the multiple condition discriminant network groups after network depth adjustment to obtain multiple of the prediction sub-networks with condition markers, and generate the bearing fault prediction model through an ensemble learning method.

2. The method for predicting bearing faults of a rotating mechanical device according to claim 1, wherein, The fault record set includes an original fault record set of historical fault records based on the target scene and a homologous fault record set of historical fault records of homologous scenes based on the target scene.

3. The method for predicting bearing faults of a rotating mechanical device according to claim 2, wherein Performing clustering division on the fault record set based on fault categories, and performing frequency dimension analysis, intensity dimension analysis, and breadth dimension analysis based on the clustering division results to obtain a multi-condition prediction reference index set, including: Taking multiple fault categories as the division target, performing clustering division on the fault record set to generate multiple condition fault record sets; Traversing the multiple condition fault record sets, counting the frequencies of fault records for frequency dimension analysis to obtain a frequency reference index set; Traversing the multiple condition fault record sets, obtaining the typical prediction accuracies of multiple fault categories for intensity dimension to obtain an intensity reference index set; Traversing the multiple condition fault record sets, evaluating the fault influence of multiple fault categories for breadth dimension analysis to obtain a breadth reference index set; Outputting the frequency reference index set, the intensity reference index set, and the breadth reference index set as the multi-condition prediction reference index set.

4. The method for predicting bearing faults of a rotating mechanical device according to claim 3, characterized in that, According to the multi - working - condition prediction reference index set, perform prediction resource allocation for multiple working conditions to obtain a prediction inclination coefficient set for multiple working conditions, where the prediction inclination coefficient set includes the sample inclination coefficient, call inclination coefficient, and scale inclination coefficient of multiple working conditions, including: Determine the sample inclination coefficient of multiple working conditions according to the frequency reference index set and the breadth reference index set; Determine the scale inclination coefficient of multiple working conditions according to the frequency reference index set and the intensity reference index set; Determine the call inclination coefficient of multiple working conditions according to the intensity reference index set and the breadth reference index set.

5. A method for predicting bearing faults of rotating mechanical equipment according to claim 1, characterized in that, Through the bearing fault prediction model, perform fault prediction based on the real - time monitoring data of the bearing to obtain a fault prediction result. After that, the method further includes: Record the fault prediction result and construct a prediction record library; According to the prediction record library, configure an accumulator to accumulate the false fault prediction results of multiple working conditions, where the false fault prediction results include false alarm fault prediction results and missed alarm fault prediction results; When the accumulated result of the accumulator is greater than or equal to a preset false alarm adjustment threshold, call the multiple false fault prediction results corresponding to the accumulated result as a false fault record set to optimize the bearing fault prediction model.

6. The method for predicting bearing faults of a rotating mechanical device according to claim 5, wherein Calling the multiple false fault prediction results corresponding to the accumulated result as a false fault record set to optimize the bearing fault prediction model includes: Obtain the fault handling information of multiple false fault prediction results; Perform prediction correction on multiple false fault prediction results according to the fault handling information to obtain an improved fault record set; Use the improved fault record set to perform enhanced supervised training on the bearing fault prediction model.

7. The method for predicting bearing faults of a rotating mechanical device according to claim 6, wherein, Using the improved fault record set to perform enhanced supervised training on the bearing fault prediction model includes: Based on the false fault record set, extract the first false fault working condition, and based on the improved fault record set, extract the corresponding first improved fault record subset; Call the multiple first - working - condition prediction sub - networks corresponding to the first false fault working condition, and perform model mutation and parameter exchange based on the multiple first - working - condition prediction sub - networks to obtain N first supplementary prediction sub - networks, where N is a positive integer greater than or equal to 2; Use the first improved fault record subset as training data to perform enhanced training on the multiple first - working - condition prediction sub - networks and N first supplementary prediction sub - networks to obtain multiple first - working - condition enhanced prediction sub - networks and evaluate the prediction performance; According to the prediction performance evaluation result, exclude the last N first - working - condition enhanced prediction sub - networks with the worst prediction performance, and synchronously retain the multiple first - working - condition enhanced prediction sub - networks to the bearing fault prediction model.

8. An electronic device, characterized in that, Including: A memory for storing computer software programs; A processor for reading and executing the computer software program, thereby implementing a method for predicting bearing faults of rotating mechanical equipment according to any one of claims 1 - 7.

9. A computer-readable storage medium, characterized in that, The computer software program is stored in the storage medium, and when the computer software program is executed by the processor, it implements a method for predicting bearing faults of rotating mechanical equipment according to any one of claims 1 - 7.

Images