Fault diagnosis model training, fault diagnosis method and device, and operating machinery

Abstract

The present invention provides a fault diagnosis model training, a fault diagnosis method and device, and an operating machine, which relates to the technical field of engineering machinery. The fault diagnosis model training method includes: based on the sample data of the operating machine, obtaining a sample data sequence corresponding to each sample moment in the sample period, and dividing each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model; obtaining a first sample label corresponding to each sample cluster, and obtaining a second sample label corresponding to the sample data based on each first sample label; based on the sample data and the second sample label, training the fault diagnosis model to obtain a trained fault diagnosis training model. The fault diagnosis model training, fault diagnosis method and device, and operating machine provided by the present invention can improve the quality of sample data and corresponding labels when the amount of sample data is small, and train a fault diagnosis model with higher calculation accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering machinery, and in particular to a fault diagnosis model training, a fault diagnosis method and device, and an operating machinery. Background Art

[0002] Operating machinery is usually an integrated mechatronic equipment including mechanical system, electrical system, control system and hydraulic system. Prompt and accurate fault diagnosis of operating machinery, especially when the operating machinery is about to or has failed, can reduce the maintenance time of the operating machinery and improve the maintenance quality, thereby ensuring that each device in the operating machinery can reliably and effectively perform its function, increase the operating time of the operating machinery in a trouble-free state, ensure the reliable operation of the operating machinery, and maximize the benefits of the operating machinery.

[0003] In the prior art, based on the historical status data of the operating machinery and artificial intelligence technology, a fault diagnosis model that can be used to describe the mapping relationship between status data and fault types can be obtained through training to achieve fault diagnosis of the operating machinery.

[0004] However, due to the complex structure of the operating machinery, the changeable working conditions and the susceptibility to interference and impact from the external environment during the operation process, the historical status data of the operating machinery usually has the characteristics of large data volume, high dimension, strong professionalism and mixed with a large amount of random noise. In the existing fault diagnosis model training method, it is difficult to completely and accurately obtain the sample data and corresponding sample labels for model training based on the historical status data of the operating machinery, which leads to the low calculation accuracy of the fault diagnosis model obtained by training, and the low accuracy of the operating machinery fault diagnosis results obtained based on the above fault diagnosis model. Summary of the invention

[0005] The present invention provides a fault diagnosis model training, a fault diagnosis method and device, and an operating machine, so as to solve the defect in the prior art that it is difficult to obtain sample data and corresponding sample labels for model training completely and accurately based on the historical status data of the operating machine, thereby resulting in low calculation accuracy of the fault diagnosis model obtained through training, and realize a fault diagnosis model with higher calculation accuracy obtained through training.

[0006] The present invention provides a fault diagnosis model training method, comprising:

[0007] Acquire sample data of the working machine, the sample data including time series data of target operating parameters within a sample period of the working machine, the target operating parameters including at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature;

[0008] Based on the sample data, a sample data sequence corresponding to each sample moment in the sample period is obtained, and each sample data sequence is divided into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model;

[0009] Obtain a first sample label corresponding to each of the sample clusters, and based on each of the first sample labels, obtain a second sample label corresponding to the sample data;

[0010] Based on the sample data and the second sample label, a fault diagnosis model is trained to obtain a trained fault diagnosis training model;

[0011] The first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; the second sample label is used to characterize the fault diagnosis result within the sample period of the operating machinery.

[0012] According to a fault diagnosis model training method provided by the present invention, the step of obtaining a first sample label corresponding to each sample cluster specifically includes:

[0013] In a case where the sample data in the sample cluster includes first sample data and second sample data, based on a state label carried by the first sample data in the sample cluster, obtaining a state label corresponding to the second sample data in the sample cluster;

[0014] Based on the state label carried by the first sample data in the sample cluster and the state label corresponding to the second sample data, obtaining a first sample label corresponding to the sample cluster;

[0015] Among them, the first sample data is sample data carrying a status label; the second sample data is sample data not carrying a status label; the status label is used to characterize at least one of the working condition, fault type and fault cause corresponding to the sample data.

[0016] According to a fault diagnosis model training method provided by the present invention, the step of obtaining a first sample label corresponding to each sample cluster specifically includes:

[0017] In a case where the sample data in the sample cluster only includes the second sample data, performing feature extraction on the second sample data in the sample cluster to obtain sample feature data corresponding to the sample cluster;

[0018] Based on the sample feature data corresponding to the sample cluster, obtaining a first sample label corresponding to the sample cluster;

[0019] The second sample data is sample data without a status tag; the status tag is used to characterize at least one of the working condition, fault type and fault cause corresponding to the sample data.

[0020] According to a fault diagnosis model training method provided by the present invention, the acquiring of a sample data sequence corresponding to each sample moment in the sample period based on the sample data specifically includes:

[0021] For each sample moment in the sample period, sample data within a preset time period before each sample moment is intercepted as a sample data sequence corresponding to each sample moment.

[0022] According to a fault diagnosis model training method provided by the present invention, the step of obtaining sample data of the operating machinery specifically includes:

[0023] Acquiring original sample data of the operating machine;

[0024] Performing data processing on the original sample data, and using the original sample data after the data processing as the sample data;

[0025] The original sample data includes the original time series data of the target operating parameters within the sample period of the operating machinery.

[0026] The present invention also provides a fault diagnosis method, comprising:

[0027] Acquire target data of the working machine, the target data including time series data of target operating parameters of the working machine within a target period, the target operating parameters including at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature;

[0028] Inputting the target data into a fault diagnosis model, and obtaining a fault diagnosis result of the working machine within the target time period output by the fault diagnosis model;

[0029] Wherein, the fault diagnosis model is trained based on any of the fault diagnosis model training methods described above.

[0030] The present invention also provides a fault diagnosis model training device, comprising:

[0031] A sample data acquisition module is used to acquire sample data of the working machine, wherein the sample data includes time series data of target operating parameters within a sample period of the working machine; the target operating parameters include at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature;

[0032] A segmentation and clustering module, configured to obtain, based on the sample data, a sample data sequence corresponding to each sample moment in the sample period, and divide each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model;

[0033] A label acquisition module, used to acquire a first sample label corresponding to each of the sample clusters, and based on each of the first sample labels, acquire a second sample label corresponding to the sample data;

[0034] A model training module, used to train a fault diagnosis model based on the sample data and the second sample label to obtain a trained fault diagnosis training model;

[0035] The first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; the second sample label is used to characterize the fault diagnosis result within the sample period of the operating machinery.

[0036] The present invention also provides a fault diagnosis device, comprising:

[0037] A target data acquisition module is used to acquire target data of the working machine; the target data includes time series data of target operating parameters of the working machine within a target period; the target operating parameters include at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature;

[0038] A fault diagnosis module, used for inputting the target data into a fault diagnosis model, and obtaining a fault diagnosis result of the working machine within the target time period output by the fault diagnosis model;

[0039] Wherein, the fault diagnosis model is trained based on the fault diagnosis model training method as described above.

[0040] A fault diagnosis device provided according to the present invention also includes: the fault diagnosis model training device as described above.

[0041] The present invention also provides a working machine, comprising: the fault diagnosis model as described above.

[0042] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the fault diagnosis model training method and / or the fault diagnosis method as described above are implemented.

[0043] 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 any of the above-described fault diagnosis model training methods and / or fault diagnosis methods.

[0044] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned fault diagnosis model training methods and / or fault diagnosis methods.

[0045] The fault diagnosis model training, fault diagnosis method and device, and operating machinery provided by the present invention obtain a sample data sequence corresponding to each sample moment in a sample period based on sample data of the operating machinery, divide each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model, obtain a first sample label corresponding to each sample cluster, and obtain a second sample label corresponding to the sample data based on each first sample label, train the fault diagnosis model based on the sample data and the corresponding second sample label to obtain a trained fault diagnosis model, can divide sample data with higher dimensions and stronger time series into sample data sequences corresponding to each sample moment, can cluster each realized data sequence based on an inverse covariance matrix and a Markov network model, can improve the quality of sample data when the sample data has a small amount of data, a small coverage, and unstable reliability, can more accurately obtain labels corresponding to sample data, can improve the efficiency of obtaining sample data and corresponding labels, can train a fault diagnosis model with higher calculation accuracy, and can further realize more accurate operating machinery fault diagnosis based on the above-mentioned fault diagnosis model. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 It is a flow chart of the fault diagnosis model training method provided by the present invention;

[0048] Figure 2 It is one of the flow charts of the fault diagnosis method provided by the present invention;

[0049] Figure 3 This is the second flow chart of the fault diagnosis method provided by the present invention;

[0050] Figure 4 It is a structural schematic diagram of the fault diagnosis model training device provided by the present invention;

[0051] Figure 5 It is a structural schematic diagram of the fault diagnosis device provided by the present invention;

[0052] Figure 6It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In the description of the invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0055] It should be noted that the traditional fault diagnosis method for operating machinery can obtain the operating status data of the operating machinery by real-time monitoring the operating parameters such as the position, operating time, engine speed, hydraulic oil temperature, etc. of the operating machinery, and perform fault diagnosis on the operating machinery based on the above operating status data and the traditional fault diagnosis model.

[0056] The above-mentioned traditional fault diagnosis model can be obtained after training by the following method: first, the historical operating status data of the operating machinery is obtained as the sample data of the operating machinery, and the fault features corresponding to the above-mentioned sample data are obtained by feature extraction of the above-mentioned sample data, so that a sample library for characterizing the historical operating status of the operating machinery can be established based on the above-mentioned fault features. Traditional feature extraction methods mainly include: time-frequency domain feature analysis, information entropy feature analysis, empirical mode decomposition, Hilbert change, and wavelet transform. Secondly, based on the above-mentioned fault features and relying on artificial experience and prior knowledge, the above-mentioned sample data can be manually labeled to obtain the sample labels corresponding to the above-mentioned sample data. Finally, a traditional fault diagnosis model can be constructed based on a classifier algorithm, and the traditional fault diagnosis model can be trained based on the above-mentioned sample data and the corresponding sample labels to obtain a trained traditional fault diagnosis model. The above-mentioned traditional fault diagnosis model can be used to describe the mapping relationship between fault features and fault types. Classifier algorithms can include neural networks, extreme learning machines, and support vector machines.

[0057] However, the traditional operating machinery fault diagnosis method has the following technical problems: First, the diagnostic strategy of the traditional operating machinery fault diagnosis method focuses on data-driven, and the operating status data of the operating machinery has the characteristics of high dimension and strong time series. After obtaining the historical operating status data of the operating machinery as the sample data of the operating machinery, the fault feature data extracted based on the above sample data is difficult to fully and accurately express the corresponding fault type. In addition, when determining the sample label corresponding to the sample data, it relies too much on manual experience and prior knowledge, which is greatly affected by the subjective factors of the technicians, and the time and labor costs are high, and the efficiency is low.

[0058] Secondly, in order to establish the mapping relationship between fault features and fault types, it is usually achieved by training a traditional fault diagnosis model based on a classifier algorithm. Since the operating status data of the operating machinery has the characteristics of high data dimension, strong time series and large noise interference, in order to avoid the problem of "dimensionality disaster" in the classifier algorithm, a large amount of sample data is often required when training the traditional fault diagnosis model, and the fault features extracted based on the above sample data need to correspond to the typical fault type. However, since the sample data collected when the fault type is known, the data quality is difficult to guarantee, and the workload of data collection is large. Moreover, even if as much sample data as possible that can correspond to the typical fault type is collected, it is difficult to ensure that all fault types are covered, resulting in the difficulty of obtaining a large amount of sample data corresponding to the typical fault type in practice. In the case of small amount of sample data, small coverage and unstable reliability, the traditional fault diagnosis model trained based on the above sample data has low classification accuracy for faults, and the effect is poor in practical applications.

[0059] In this regard, the present invention provides a fault diagnosis model training method and a fault diagnosis method. Based on the fault diagnosis model training method provided by the present invention, a fault diagnosis model with high calculation accuracy can be trained when the sample data has a small amount of data, a small coverage area, and unstable reliability. Based on the fault diagnosis method provided by the present invention, a more accurate fault diagnosis can be performed on the operating machinery based on the above-mentioned fault diagnosis model.

[0060] Figure 1 Schematic diagram of the process of the fault diagnosis model training method provided by the embodiment of the present invention. Figure 1 The fault diagnosis model training method of the present invention is described. Figure 1 As shown, the method includes: step 101, obtaining sample data of the working machinery, the sample data including time series data of target operating parameters within a sample period of the working machinery, the target operating parameters including at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature.

[0061] It should be noted that the fault diagnosis model can be used for fault diagnosis of working machinery. Based on the fault diagnosis model training method provided by the present invention, the above-mentioned fault diagnosis model can be trained to obtain a trained fault diagnosis model, and the target data of the working machinery can be input into the above-mentioned trained fault diagnosis model to obtain the fault diagnosis result of the working machinery within the target period output by the above-mentioned trained fault diagnosis model. Among them, the target data of the working machinery can include at least one of the time series data of the engine speed within the target period of the working machinery, the time series data of the cooling water temperature, the time series data of the hydraulic oil pump pressure, and the time series data of the hydraulic oil temperature.

[0062] It should be noted that the operating machinery in the embodiments of the present invention refers to mechanical equipment that can perform transportation operations and mechanized construction, such as various types of household cars and engineering vehicles such as excavators and concrete pump trucks.

[0063] During the operation of the operating machinery, various sensors can be used to monitor the operating machinery and obtain time series data that can be used to describe the operating status of the operating machinery.

[0064] It is understandable that due to the wide variety of operating machines and their complex structures, some parts of the operating machines and even the operating machines themselves can perform different forms of movement when the operating machines are operating, and the number of sensors monitoring the operating machines is large and the types are diverse. Therefore, in the embodiment of the present invention, based on the physical mechanism of the operation of the operating machine, combined with the type and layout of the sensor, a sensor that can characterize the typical fault state of the operating machine can be selected from the above-mentioned various sensors to determine as the target sensor.

[0065] After determining the target sensor, the operating parameters corresponding to the target sensor can be determined as the target operating parameters. Based on the above target sensor, the time series data of the target operating parameters of the operating machinery can be obtained. The time series data of the target operating parameters collected by the above target sensor during the sample period can better describe and characterize the fault state of the operating machinery during the sample period, thereby better meeting the training requirements of the fault diagnosis model.

[0066] Optionally, the main systems in the working machine include but are not limited to at least one of a driving system, a cooling system and a hydraulic system. In order to characterize the fault state of the main systems of the working machine, one or more of an engine speed sensor, a cooling water temperature sensor, a hydraulic oil pump pressure sensor and a hydraulic oil temperature sensor may be determined as the target sensor, that is, at least one of the engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature corresponding to the target sensor may be determined as the target operating parameter.

[0067] Based on the above target sensors, the time series data of the target operating parameters within the sample period of the operating machinery can be obtained, and the time series data of the above target operating parameters can be arranged in the order of acquisition time to obtain the sample data of the operating machinery.

[0068] Optionally, based on the above-mentioned target sensor, the original time series data of the target operating parameters within the sample period of the operating machinery can be obtained as the original sample data, and after arranging the above-mentioned original sample data in the order of acquisition time, the above-mentioned sorted original sample data is used as the sample data of the operating machinery; or, the above-mentioned original sample data can also be processed, and the above-mentioned original sample data after data processing can be arranged in the order of acquisition time, and the above-mentioned original sample data after data processing can be used as the sample data of the operating machinery.

[0069] It should be noted that the data processing of the original sample data may include but is not limited to eliminating incomplete and low-accuracy data in the original sample data, removing noise data in the original sample data, etc.

[0070] Step 102: Based on the sample data, a sample data sequence corresponding to each sample time in the sample period is obtained, and each sample data sequence is divided into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model.

[0071] Specifically, after obtaining the sample data within the sample period of the operating machinery, the sample data sequence corresponding to each sample moment within the sample period can be obtained in a variety of ways based on the above sample data. For example, the sample data within the first preset time length before each sample moment can be intercepted as the sample data sequence corresponding to each sample moment; or the sample data within the second preset time length with each sample moment as the midpoint of time can be intercepted as the sample data sequence corresponding to each sample moment.

[0072] It should be noted that the first preset time length and the second preset time length may be the same or different.

[0073] After obtaining the sample data sequence corresponding to each sample moment in the sample period, each sample data sequence can be divided into one or more sample clusters based on the sparse inverse covariance matrix and the Markov network model through numerical calculation, mathematical statistics, etc. Wherein, in the case where any sample cluster includes multiple sample data sequences, the similarity of the multiple sample data sequences is relatively high.

[0074] It should be noted that the time interval between any two adjacent sample moments in the sample period may be the same or different. When the time interval between any two adjacent sample moments is the same, the time interval may be predetermined based on actual conditions. When the time interval between any two adjacent sample moments is different, each sample moment in the sample period may be predetermined based on actual conditions.

[0075] It is understandable that, for any sample moment within the sample period, the sample data corresponding to the sample moment can only represent the operating status of the operating machinery at the sample moment. The sample data sequence corresponding to the sample moment includes the sample data before and after the sample moment, which can more completely characterize the continuous operating status of the operating machinery at the sample moment. Therefore, in the embodiment of the present invention, the sample data corresponding to each sample moment is not directly clustered, but the sample data sequence corresponding to each sample moment is clustered, and the sample data adjacent in time series are divided into the same sample data sequence, so that the time consistency of the sample data sequence can be achieved.

[0076] It should be noted that, since the classifier algorithm based on the traditional fault diagnosis model training method usually relies on the distance metric, such as Dynamic Time Warping (DTW), the traditional fault diagnosis model training method focuses more on matching sample data rather than finding similarities in sample data. Therefore, it is difficult to accurately and efficiently obtain the similarities (or differences) between different data sources or the same data source in different time intervals based on the traditional fault diagnosis model training method, that is, it is difficult for the traditional fault diagnosis model training method to accurately and efficiently obtain the similarities (or differences) between any two of the above sample data sequences, which in turn makes it difficult for the traditional fault diagnosis model training method to accurately and efficiently divide the above sample data sequences into one or more sample clusters, and extract association rules from the above sample data sequences to discover knowledge.

[0077] The inverse covariance matrix is ​​the inverse matrix of the covariance matrix, which is proportional to the partial correlation matrix. The inverse covariance matrix can give the independence relationship between the data. The sparse inverse covariance matrix refers to the sparse inverse covariance matrix. When the sample size is small, the sparse inverse covariance matrix can restore the off-diagonal structure.

[0078] The Markov network model (MRF), also known as the Markov random field dependency network model, is a graph-based dependency network model. Any edge in the Markov network model can represent the partial correlation between two variables and can be used to characterize the dependency relationship between the two variables.

[0079] In the embodiment of the present invention, each sample data sequence may be divided into one or more sample clusters based on a sparse inverse covariance matrix, so that the sample data may be decomposed into a clearer sequential timeline.

[0080] For any sample cluster, the Markov network model can be used to characterize the characteristics of the sample cluster. The Markov network corresponding to the sample cluster includes multiple layers. There are edges in any layer and between different layers. The above edges can represent the relationship between the time series data collected by the same target sensor at different times, and can also represent the relationship between the time series collected by different target sensors at the same time or at different times. The Markov network corresponding to the sample cluster can characterize the non-time-varying correlation structure of each sample data sequence in the sample cluster.

[0081] In an embodiment of the present invention, in order to enable the fault diagnosis model to learn the Markov network corresponding to each sample cluster, a sparse inverse covariance matrix can be estimated to learn the Markov network corresponding to each sample cluster, thereby ensuring that the clustering of the fault diagnosis model is time-invariant, that is, when each sample data sequence is divided into one or more sample clusters, it does not depend on the exact starting position of each sample data sequence, but depends on the structural state of each sample data sequence.

[0082] Based on the sparse inverse covariance matrix and Markov network, each sample data sequence is divided into one or more sample clusters. The key sample data in any sample cluster and the relationship between the sample data can be described, so as to better explain the sample cluster, provide better interpretability for the division of sample clusters, prevent overfitting, and allow the discovery of pattern types that cannot be discovered by other means.

[0083] Based on the sparse inverse covariance matrix and the Markov network, each sample data sequence can be more accurately divided into one or more sample clusters when the sample data is small.

[0084] It should be noted that the number of layers of the Markov network corresponding to the sample cluster has a corresponding relationship with the first preset time length or the second preset time length.

[0085] Step 103: Obtain a first sample label corresponding to each sample cluster, and based on each first sample label, obtain a second sample label corresponding to the sample data.

[0086] Among them, the first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; the second sample label is used to characterize the fault diagnosis result within the operating machinery sample period.

[0087] Specifically, after each sample data sequence is divided into one or more sample clusters, the first sample label corresponding to each sample cluster can be obtained in a variety of ways. For example: for any sample cluster, the fault type corresponding to the sample data can be determined based on the sample data carrying the state label in the sample cluster, and the fault type corresponding to the sample data can be inferred through numerical calculation, mathematical statistics, etc., based on the fault type corresponding to the sample data, to obtain the fault type corresponding to all the sample data in the sample cluster, and the fault type corresponding to all the sample data in the sample cluster can be used as the first sample label corresponding to the sample cluster; or, feature extraction can be performed on all the sample data in the sample cluster, and the first sample label corresponding to the sample cluster can be obtained based on the result of the feature extraction.

[0088] It should be noted that the first sample label corresponding to the sample cluster can be used to characterize the fault diagnosis result corresponding to the sample cluster. The second sample label corresponding to the sample data of the working machine can be used to characterize the fault diagnosis result within the sample period of the working machine.

[0089] The above fault diagnosis results may include the fault type and / or fault cause. The above fault type may include but is not limited to drive system fault, cooling system fault and hydraulic system fault. The above fault cause may include but is not limited to engine abnormality, excessive impurities in hydraulic oil and blockage of heat dissipation device.

[0090] It should be noted that the status labels carried by the sample data may include but are not limited to the working conditions corresponding to the sample data (such as earthwork conditions, rockwork conditions and crushing conditions of the excavator), fault types and at least one of the fault causes.

[0091] It is understandable that different sample clusters correspond to different Markov networks. For example: if the first sample label corresponding to the first sample cluster includes a drive device failure, and the first sample label corresponding to the second sample cluster includes a cooling system failure, then the dependency results between the time series data of the engine speed and the time series data of the cooling water temperature in the Markov networks corresponding to the above two sample clusters are very different. The Markov network corresponding to the first sample cluster can describe how the engine speed at time t affects the cooling water temperature at time t+1. The relationship represented by the Markov network is more comprehensive and flexible than a simple correlation, which can greatly enhance the interpretability of the sample cluster and can more accurately describe the key sample data in the sample cluster and the relationship between each sample data.

[0092] After obtaining the first sample label corresponding to each sample cluster, the second sample label corresponding to the sample data within the sample period of the operating machinery can be obtained in a variety of ways based on the first sample label corresponding to each sample cluster. For example, the first sample label corresponding to each sample cluster can be listed as the above-mentioned second sample label; or, the top three fault types with the largest number included in each first sample label can be obtained as the above-mentioned second sample label through mathematical statistics.

[0093] Step 104: Train the fault diagnosis model based on the sample data and the second sample label to obtain a trained fault diagnosis training model.

[0094] Specifically, after obtaining the second sample label corresponding to the sample data, the fault diagnosis model can be trained based on the sample data and the second sample label to obtain a trained fault diagnosis model.

[0095] The embodiment of the present invention obtains a sample data sequence corresponding to each sample moment in a sample period based on sample data of an operating machine, divides each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model, obtains a first sample label corresponding to each sample cluster, and obtains a second sample label corresponding to the sample data based on each first sample label, trains a fault diagnosis model based on the sample data and the corresponding second sample label, and obtains a trained fault diagnosis model. The model can divide sample data with higher dimensions and stronger temporal sequence into a sample data sequence corresponding to each sample moment, and cluster each data sequence based on an inverse covariance matrix and a Markov network model, so as to improve the quality of sample data, more accurately obtain labels corresponding to sample data, and improve the efficiency of obtaining sample data and corresponding labels when the sample data has a small amount of data, a small coverage, and unstable reliability, and train a fault diagnosis model with higher calculation accuracy, so as to realize more accurate operating machine fault diagnosis based on the above fault diagnosis model.

[0096] Based on the contents of the above embodiments, obtaining the first sample label corresponding to each sample cluster specifically includes: when the sample data in the sample cluster includes the first sample data and the second sample data, based on the status label carried by the first sample data in the sample cluster, obtaining the status label corresponding to the second sample data in the sample cluster.

[0097] Among them, the first sample data is sample data carrying a status label; the second sample data is sample data not carrying a status label; the status label is used to characterize at least one of the working condition, fault type and fault cause corresponding to the sample data.

[0098] It should be noted that when manually labeling sample data based on manual experience and prior knowledge, it may be impossible to manually label all sample data due to difficulty in determining the specific fault type, omissions, etc. However, in the case where some sample data lack corresponding manual labels, it is difficult for traditional fault diagnosis model training methods to accurately obtain the second sample labels corresponding to the above sample data. In an embodiment of the present invention, the second sample labels corresponding to the above sample data can be obtained when some sample data carry state labels and some sample data do not carry state labels.

[0099] It should be noted that the status labels carried by the sample data may be determined based on manual annotation.

[0100] Specifically, the sample data carrying the state label may be used as the first sample data, and the sample data not carrying the state label may be used as the second sample data.

[0101] After dividing each sample data sequence into one or more sample clusters, for any sample cluster, if the sample data in the sample cluster includes the first sample data and the second sample data, it means that some sample data in the sample cluster lacks the corresponding manual annotation. In the embodiment of the present invention, based on the state label carried by the first sample data in the sample cluster, the state label corresponding to the second sample data in the sample cluster can be inferred by numerical calculation, mathematical statistics, etc.

[0102] For example: for any sample cluster, the fault type corresponding to the first sample data in the sample cluster can be determined based on the status label carried by the first sample data in the sample cluster, and the fault type corresponding to the first sample data can be inferred through numerical calculation, mathematical statistics, etc., as the status label corresponding to the second sample data in the sample cluster.

[0103] Based on the state label carried by the first sample data in the sample cluster and the state label corresponding to the second sample data, a first sample label corresponding to the sample cluster is obtained.

[0104] Specifically, after obtaining the state label corresponding to the second sample data in any sample cluster, the first sample label corresponding to the sample cluster can be obtained based on the state label carried by the first sample data in the sample cluster and the state label corresponding to the second sample data through numerical calculation, mathematical statistics, etc. For example: based on the state label carried by the first sample data in the sample cluster and the state label corresponding to the second sample data, the corresponding fault type in the sample cluster can be obtained, and the fault type corresponding to the sample cluster can be used as the first sample label corresponding to the sample cluster.

[0105] In an embodiment of the present invention, when a sample cluster includes first sample data and second sample data, a status label corresponding to the second sample data in the sample cluster is obtained based on a status label carried by the first sample data in the sample cluster, and a first sample label corresponding to the sample cluster is obtained based on a status label carried by the first sample data in the sample cluster and a status label corresponding to the second sample data. This can more accurately and efficiently obtain labels corresponding to sample data when some sample data lack manual annotation.

[0106] Based on the contents of the above embodiments, obtaining the first sample label corresponding to each sample cluster specifically includes: when the sample data in the sample cluster only includes the second sample data, performing feature extraction on the second sample data in the sample cluster to obtain sample feature data corresponding to the sample cluster.

[0107] Specifically, when manually labeling sample data based on human experience and prior knowledge, a lot of time and labor costs are required, and manual labeling is greatly affected by the subjective factors of technicians. Therefore, after obtaining sample data in the embodiment of the present invention, no manual labeling is required, that is, all sample data do not carry state labels.

[0108] After dividing each sample data sequence into one or more sample clusters, for any sample cluster, if the sample data in the sample cluster only includes the second sample data, it means that all sample data in the sample cluster lacks the corresponding manual annotation. In the embodiment of the present invention, feature extraction can be performed on the second sample data in the sample cluster to obtain sample feature data corresponding to the sample cluster.

[0109] Optionally, feature extraction of the second sample data in the sample cluster may include obtaining the maximum, minimum, average, slope and variance of the second sample data in the sample cluster as sample feature data corresponding to the sample cluster by numerical calculation.

[0110] Based on the sample feature data corresponding to the sample cluster, a first sample label corresponding to the sample cluster is obtained.

[0111] Specifically, after obtaining the sample feature data corresponding to any sample cluster, the first sample label corresponding to the sample cluster may be obtained based on the sample feature data corresponding to the sample cluster and prior knowledge.

[0112] In the embodiment of the present invention, when the sample cluster only includes the second sample data, feature extraction is performed on the second sample data in the sample cluster to obtain sample feature data corresponding to the sample cluster, and based on the sample feature data corresponding to the sample data, a first sample label corresponding to the sample data is obtained. In the case where all sample data lack manual labeling, the label corresponding to the sample data can be obtained more accurately, thereby avoiding the label corresponding to the sample data from being affected by subjective factors during manual labeling, and further improving the efficiency of obtaining the label corresponding to the sample data.

[0113] Based on the contents of the above embodiments and based on the sample data, a sample data sequence corresponding to each sample moment in the sample time period is obtained, specifically including: for each sample moment in the sample time period, the sample data within a preset time length before each sample moment is intercepted as the sample data sequence corresponding to each sample moment.

[0114] Specifically, after obtaining the sample data of the operating machine, the sample data can be expressed as:

[0115] x=[x1,x2,x3,…,x T ]

[0116] Where x represents the sample data; T represents the length of the sample period; x i represents the sample data at the i-th sample time in the sample period, i = 1, 2, ..., T. In the case where the number of target sensors is n, x i is an n-dimensional vector, where n is a positive integer greater than 0.

[0117] In order to facilitate the analysis of the correlation before and after each sample moment, in the embodiment of the present invention, each sample moment is taken as a reference, and the sample data within a preset time length before each sample moment is intercepted as the sample data sequence corresponding to each sample moment. The sample data sequence corresponding to the i-th sample moment can be expressed as:

[0118] X i =[x i-w+1 ,x i-w ,x i-w-1 ,…,x i-1 ,x i ]

[0119] Wherein, w represents the above-mentioned preset duration.

[0120] It should be noted that the longer the above preset time length is, the farther the distance across the time boundary of the sample data sequence corresponding to each sample moment will be, which may result in low classification accuracy of the sample data at the segmentation boundary. Therefore, the above preset time length needs to be kept relatively short.

[0121] In the embodiment of the present invention, the preset duration can be determined by cross-validation according to the working conditions of the working machine in the sample period. For example, the time period required for the working machine to complete different working actions is different, and the average cycle length required for the working action performed by the working machine in the sample period can be used as the initial value of the preset duration. In addition, the sparse inverse covariance matrix has good robustness for determining the preset duration.

[0122] The embodiment of the present invention intercepts sample data within a preset time length before each sample moment in the sample period as the sample data sequence corresponding to each sample moment, and divides the sample data into a clearer sequential timeline by segmenting the sample data, more completely expressing the continuous data state, thereby improving the accuracy of clustering.

[0123] Based on the contents of the above embodiments, obtaining sample data of the operating machine specifically includes: obtaining original sample data of the operating machine.

[0124] Among them, the original sample data includes the original time series data of the target operating parameters within the sample period of the operating machinery.

[0125] Specifically, based on the above-mentioned target sensor, the original time series data of the target operating parameters within the sample period of the working machinery can be obtained.

[0126] The original sample data is processed, and the processed original sample data is used as sample data.

[0127] Specifically, after acquiring the above-mentioned original sample data, data processing may be performed on the above-mentioned original sample data, and the original sample data after data processing may be used as sample data within a sample period of the operating machinery.

[0128] Optionally, data processing of the original sample data may include but is not limited to: eliminating incomplete and low-accuracy data in the original sample data, removing noise data in the original sample data, and sorting the data in chronological order of acquisition time, etc., so as to ensure the data quality of the obtained sample data.

[0129] Optionally, the data processing performed on the original sample data may further include manually labeling part of the original sample data to obtain part of the original sample data carrying state labels.

[0130] It should be noted that when it is difficult to accurately manually label certain original sample data, there is no need to label the above-mentioned original sample data.

[0131] The embodiment of the present invention obtains the original sample data of the operating machinery, processes the original sample data, and uses the processed original sample data as sample data, which can ensure the data quality of the sample data and further improve the calculation accuracy of the fault diagnosis model obtained by training.

[0132] Figure 2 This is one of the flow charts of the fault diagnosis method provided by the present invention. Figure 2 The fault diagnosis method of the present invention is described. Figure 2 As shown, the method includes: step 201, obtaining target data of the working machinery; the target data includes time series data of target operating parameters of the working machinery within a target period; the target operating parameters include at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature.

[0133] Specifically, based on the above-mentioned target sensor, target data including that of the working machinery can be acquired; wherein the above-mentioned target data includes one or more of the timing data of the engine speed within the target period of the working machinery, the timing data of the cooling water temperature, the timing data of the hydraulic oil pump pressure and the timing data of the hydraulic oil temperature.

[0134] Step 202: Input the target data into the fault diagnosis model, and obtain the fault diagnosis result of the working machine within the target time period output by the fault diagnosis model.

[0135] The fault diagnosis model is obtained by training based on any of the above-mentioned fault diagnosis model training methods.

[0136] It should be noted that the fault diagnosis model can be trained based on the fault diagnosis model training method in the above embodiments to obtain a trained fault diagnosis model. The specific process of training the fault diagnosis model can refer to the contents of the above embodiments, and will not be repeated in the embodiments of the present invention.

[0137] After acquiring the target data of the operating machine, the target data may be input into the trained fault diagnosis model, and the fault diagnosis result of the operating machine within the target time period output by the trained fault diagnosis model may be acquired.

[0138] Optionally, the above-mentioned fault diagnosis result may include the fault type of the operating machinery that fails within the target time period.

[0139] It should be noted that the target data of the working machine in the embodiment of the present invention is of the same type as the target operating parameter in the sample data used to train the fault diagnosis model.

[0140] After acquiring the target data of the working machinery, the embodiment of the present invention inputs the target data into a trained fault diagnosis model to obtain the fault diagnosis result of the working machinery within the target time period output by the trained fault diagnosis model. The target data includes at least one of the timing data of the engine speed of the working machinery within the target time period, the timing data of the cooling water temperature, the timing data of the hydraulic oil pump pressure and the timing data of the hydraulic oil temperature. Based on the fault diagnosis model with higher calculation accuracy, more accurate fault diagnosis can be performed on the working machinery.

[0141] In order to facilitate the understanding of the fault diagnosis method provided by the present invention, the fault diagnosis method provided by the present invention is described below by using an example. Figure 3 FIG. 2 is a flow chart of the fault diagnosis method provided by the present invention. Figure 3 As shown, the method includes: determining a target sensor based on the physical mechanism of the operation of the operating machine, and using the target sensor to obtain original time series data within a sample period of the operating machine as original sample data.

[0142] The original sample data is processed to obtain the sample data of the operating machinery. Based on the characteristics of the working mode within the sample period of the operating machinery, the preset duration is determined through cross-validation, and the sample data is segmented based on the preset duration to obtain the sample data sequence corresponding to each sample moment within the sample period.

[0143] Based on the sparse inverse covariance matrix and the Markov network model, each sample data sequence is clustered and divided into at least one sample cluster.

[0144] A first sample label corresponding to each sample cluster is obtained, and based on each first sample label, a second sample label corresponding to the sample data is obtained.

[0145] The fault diagnosis model is trained based on the sample data and the corresponding second sample labels to obtain a trained fault diagnosis model.

[0146] After acquiring the target data of the operating machinery, the target data is input into a trained fault diagnosis model to obtain the fault diagnosis result of the operating machinery within the target period output by the trained fault diagnosis model.

[0147] Figure 4 Schematic diagram of the structure of the fault diagnosis model training device provided by the present invention. Figure 4 The fault diagnosis model training device provided by the present invention is described. The fault diagnosis model training device described below and the fault diagnosis model training method provided by the present invention described above can be referred to each other. Figure 4As shown, the device includes: a sample data acquisition module 401, a segmentation and clustering module 402, a label acquisition module 403 and a model training module 404.

[0148] The sample data acquisition module 401 is used to acquire sample data of the operating machinery, the sample data including time series data of target operating parameters within a sample period of the operating machinery; the target operating parameters including at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature.

[0149] The segmentation and clustering module 402 is used to obtain a sample data sequence corresponding to each sample time in a sample period based on the sample data, and divide each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model.

[0150] The label acquisition module 403 is used to acquire the first sample label corresponding to each sample cluster, and acquire the second sample label corresponding to the sample data based on each first sample label.

[0151] The model training module 404 is used to train the fault diagnosis model based on the sample data and the second sample label to obtain a trained fault diagnosis training model.

[0152] Among them, the first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; the second sample label is used to characterize the fault diagnosis result within the operating machinery sample period.

[0153] Specifically, the sample data acquisition module 401, the segmentation and clustering module 402, the label acquisition module 403 and the model training module 404 are electrically connected.

[0154] Optionally, the label acquisition module 403 can be specifically used to obtain a status label corresponding to the second sample data in the sample cluster based on a status label carried by the first sample data in the sample cluster when the sample data in the sample cluster includes first sample data and second sample data; and to obtain a first sample label corresponding to the sample cluster based on the status label carried by the first sample data in the sample cluster and the status label corresponding to the second sample data; wherein the first sample data is sample data carrying a status label; and the second sample data is sample data not carrying a status label; and the status label is used to characterize at least one of the operating condition, fault type and fault cause corresponding to the sample data.

[0155] Optionally, the label acquisition module 403 can also be specifically used to perform feature extraction on the second sample data in the sample cluster to obtain sample feature data corresponding to the sample cluster when the sample data in the sample cluster only includes the second sample data; and obtain the first sample label corresponding to the sample cluster based on the sample feature data corresponding to the sample cluster.

[0156] Optionally, the segmentation and clustering module 402 includes a data segmentation unit.

[0157] The data segmentation unit may be used to intercept the sample data within a preset time period before each sample moment in the sample period as the sample data sequence corresponding to each sample moment.

[0158] Optionally, the sample data acquisition module 401 can be specifically used to acquire original sample data of the operating machinery; perform data processing on the original sample data, and use the processed original sample data as sample data; wherein the original sample data includes original time series data of target operating parameters within a sample period of the operating machinery.

[0159] The embodiment of the present invention obtains a sample data sequence corresponding to each sample moment in a sample period based on sample data of an operating machine, divides each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model, obtains a first sample label corresponding to each sample cluster, and obtains a second sample label corresponding to the sample data based on each first sample label, trains a fault diagnosis model based on the sample data and the corresponding second sample label, and obtains a trained fault diagnosis model. The model can divide sample data with higher dimensions and stronger temporal sequence into a sample data sequence corresponding to each sample moment, and cluster each data sequence based on an inverse covariance matrix and a Markov network model, so as to improve the quality of sample data, more accurately obtain labels corresponding to sample data, and improve the efficiency of obtaining sample data and corresponding labels when the sample data has a small amount of data, a small coverage, and unstable reliability, and train a fault diagnosis model with higher calculation accuracy, so as to realize more accurate operating machine fault diagnosis based on the above fault diagnosis model.

[0160] Figure 5 This is a schematic diagram of the structure of the fault diagnosis device provided by the present invention. Figure 5 The fault diagnosis device provided by the present invention is described. The fault diagnosis device described below and the fault diagnosis method provided by the present invention described above can be referred to each other. Figure 5 As shown, the device includes: a target data acquisition module 501 and a fault diagnosis module 502.

[0161] The target data acquisition module 501 is used to acquire the target data of the working machine; the target data includes the time series data of the target operating parameters of the working machine within the target period; the target operating parameters include at least one of the engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature.

[0162] The fault diagnosis module 502 is used to input the target data into the fault diagnosis model, and obtain the fault diagnosis result of the working machine within the target time period output by the fault diagnosis model.

[0163] The fault diagnosis model is obtained by training based on the fault diagnosis model training method as described above.

[0164] Specifically, the target data acquisition module 501 and the fault diagnosis module 502 are electrically connected.

[0165] Based on the contents of the above embodiments, the fault diagnosis device further includes: the fault diagnosis model training device as described above.

[0166] Specifically, the fault diagnosis device includes the fault diagnosis model training device as described above, and can train the fault diagnosis model based on the fault diagnosis model training device to obtain a trained fault diagnosis training model.

[0167] After acquiring the target data of the working machinery, the embodiment of the present invention inputs the target data into a trained fault diagnosis model to obtain the fault diagnosis result of the working machinery within the target time period output by the trained fault diagnosis model. The target data includes at least one of the timing data of the engine speed of the working machinery within the target time period, the timing data of the cooling water temperature, the timing data of the hydraulic oil pump pressure and the timing data of the hydraulic oil temperature. Based on the fault diagnosis model with higher calculation accuracy, more accurate fault diagnosis can be performed on the working machinery.

[0168] Based on the contents of the above embodiments, a working machine includes: the fault diagnosis device as described above.

[0169] Specifically, the working machine includes the fault diagnosis device as described above, and can perform more accurate fault diagnosis on the working machine based on a fault diagnosis model with higher calculation accuracy.

[0170] The training method of the above fault diagnosis model and the specific process of fault diagnosis for the operating machine can be found in the contents of the above embodiments, which will not be described in detail in the embodiments of the present invention.

[0171] The working machinery in the embodiment of the present invention includes a fault diagnosis device. After acquiring target data of the working machinery, the fault diagnosis device inputs the target data into a trained fault diagnosis model to obtain the fault diagnosis result of the working machinery within a target time period output by the trained fault diagnosis model. The target data includes at least one of the timing data of the engine speed of the working machinery within the target time period, the timing data of the cooling water temperature, the timing data of the hydraulic oil pump pressure and the timing data of the hydraulic oil temperature. Based on a fault diagnosis model with higher calculation accuracy, more accurate fault diagnosis can be performed on the working machinery.

[0172] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor (processor) 610, a communication interface (Communications Interface) 620, a memory (memory) 630 and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call the logic instructions in the memory 630 to execute the fault diagnosis model training method and / or the fault diagnosis method, and the fault diagnosis model training method includes: obtaining sample data of the working machinery, the sample data includes the time series data of the target operating parameters within the sample period of the working machinery, and the target operating parameters include at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; based on the sample data, obtaining the sample data sequence corresponding to each sample moment within the sample period, and dividing each sample data sequence into at least one sample cluster based on the sparse inverse covariance matrix and the Markov network model; obtaining the first sample label corresponding to each sample cluster, and based on each first sample label, obtaining the second sample label corresponding to the sample data; based on the sample data and the second sample label, training the fault diagnosis model to obtain a trained fault diagnosis training model; wherein the first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; the second sample label is used to characterize the fault diagnosis result within the sample period of the working machinery. The fault diagnosis method includes: acquiring target data of the operating machinery; the target data includes time series data of target operating parameters of the operating machinery within a target time period; the target operating parameters include at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; inputting the target data into a fault diagnosis model to obtain the fault diagnosis result of the operating machinery within the target time period output by the fault diagnosis model; wherein the fault diagnosis model is trained based on the fault diagnosis model training method as described above.

[0173] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0174] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the fault diagnosis model training method and / or fault diagnosis method provided by the above methods. The fault diagnosis model training method includes: obtaining sample data of the operating machinery, the sample data including time series data of the target operating parameters within a sample period of the operating machinery, the target operating parameters including at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; based on the sample data, obtaining a sample data sequence corresponding to each sample moment within the sample period, and dividing each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model; obtaining a first sample label corresponding to each sample cluster, and based on each first sample label, obtaining a second sample label corresponding to the sample data; based on the sample data and the second sample label, training the fault diagnosis model to obtain a trained fault diagnosis training model; wherein the first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; and the second sample label is used to characterize the fault diagnosis result within the sample period of the operating machinery. The fault diagnosis method includes: acquiring target data of the operating machinery; the target data includes time series data of target operating parameters of the operating machinery within a target time period; the target operating parameters include at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; inputting the target data into a fault diagnosis model to obtain the fault diagnosis result of the operating machinery within the target time period output by the fault diagnosis model; wherein the fault diagnosis model is trained based on the fault diagnosis model training method as described above.

[0175] On the other hand, 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 execute the fault diagnosis model training method and / or fault diagnosis method provided by the above-mentioned methods, the fault diagnosis model training method comprising: obtaining sample data of the operating machinery, the sample data comprising time series data of target operating parameters within a sample period of the operating machinery, the target operating parameters comprising at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; based on the sample data, obtaining a sample data sequence corresponding to each sample moment within the sample period, and dividing each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model; obtaining a first sample label corresponding to each sample cluster, and based on each first sample label, obtaining a second sample label corresponding to the sample data; training the fault diagnosis model based on the sample data and the second sample label to obtain a trained fault diagnosis training model; wherein the first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; and the second sample label is used to characterize the fault diagnosis result within the sample period of the operating machinery. The fault diagnosis method includes: acquiring target data of the operating machinery; the target data includes time series data of target operating parameters of the operating machinery within a target time period; the target operating parameters include at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; inputting the target data into a fault diagnosis model to obtain the fault diagnosis result of the operating machinery within the target time period output by the fault diagnosis model; wherein the fault diagnosis model is trained based on the fault diagnosis model training method as described above.

[0176] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0177] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault diagnosis model training method, characterized in that: include: Acquire sample data of the working machine, the sample data including time series data of target operating parameters within a sample period of the working machine, the target operating parameters including at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; Based on the sample data, a sample data sequence corresponding to each sample moment in the sample period is obtained, and each sample data sequence is divided into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model; Obtain a first sample label corresponding to each of the sample clusters, and based on each of the first sample labels, obtain a second sample label corresponding to the sample data; Based on the sample data and the second sample label, a fault diagnosis model is trained to obtain a trained fault diagnosis training model; Wherein, the first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; The second sample label is used to characterize the fault diagnosis result within a sample period of the working machine.

2. The fault diagnosis model training method according to claim 1, characterized in that: The obtaining of the first sample label corresponding to each of the sample clusters specifically includes: In a case where the sample data in the sample cluster includes first sample data and second sample data, based on a state label carried by the first sample data in the sample cluster, obtaining a state label corresponding to the second sample data in the sample cluster; Based on the state label carried by the first sample data in the sample cluster and the state label corresponding to the second sample data, obtaining a first sample label corresponding to the sample cluster; Among them, the first sample data is sample data carrying a status label; the second sample data is sample data not carrying a status label; the status label is used to characterize at least one of the working condition, fault type and fault cause corresponding to the sample data.

3. The fault diagnosis model training method according to claim 1, characterized in that: The obtaining of the first sample label corresponding to each of the sample clusters specifically includes: In a case where the sample data in the sample cluster only includes the second sample data, performing feature extraction on the second sample data in the sample cluster to obtain sample feature data corresponding to the sample cluster; Based on the sample feature data corresponding to the sample cluster, obtaining a first sample label corresponding to the sample cluster; The second sample data is sample data without a status tag; the status tag is used to characterize at least one of the working condition, fault type and fault cause corresponding to the sample data.

4. The fault diagnosis model training method according to claim 1, characterized in that: The acquiring, based on the sample data, a sample data sequence corresponding to each sample moment in the sample period specifically includes: For each sample moment in the sample period, sample data within a preset time period before each sample moment is intercepted as a sample data sequence corresponding to each sample moment.

5. The fault diagnosis model training method according to any one of claims 1 to 4, characterized in that: The obtaining of sample data of the operating machinery specifically includes: Acquiring original sample data of the operating machine; Performing data processing on the original sample data, and using the original sample data after the data processing as the sample data; The original sample data includes the original time series data of the target operating parameters within the sample period of the operating machinery.

6. A fault diagnosis method, characterized in that: include: Acquire target data of the working machine, the target data including time series data of target operating parameters of the working machine within a target period, the target operating parameters including at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; Inputting the target data into a fault diagnosis model, and obtaining a fault diagnosis result of the working machine within the target time period output by the fault diagnosis model; Wherein, the fault diagnosis model is trained based on the fault diagnosis model training method according to any one of claims 1 to 5.

7. A fault diagnosis model training device, characterized in that: include: A sample data acquisition module, used to acquire sample data of the operating machine, wherein the sample data includes time series data of target operating parameters within a sample period of the operating machine; The target operating parameter includes at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; A segmentation and clustering module, configured to obtain, based on the sample data, a sample data sequence corresponding to each sample moment in the sample period, and divide each sample data sequence into at least one sample cluster based on a sparse inverse covariance matrix and a Markov network model; A label acquisition module, used to acquire a first sample label corresponding to each of the sample clusters, and based on each of the first sample labels, acquire a second sample label corresponding to the sample data; A model training module, used to train a fault diagnosis model based on the sample data and the second sample label to obtain a trained fault diagnosis training model; Wherein, the first sample label is used to characterize the fault diagnosis result corresponding to the corresponding sample cluster; The second sample label is used to characterize the fault diagnosis result within a sample period of the working machine.

8. A fault diagnosis device, characterized in that: include: A target data acquisition module, used to acquire target data of the operating machinery; The target data includes time series data of target operating parameters of the operating machine within a target period of time; The target operating parameter includes at least one of engine speed, cooling water temperature, hydraulic oil pump pressure and hydraulic oil temperature; A fault diagnosis module, used for inputting the target data into a fault diagnosis model, and obtaining a fault diagnosis result of the working machine within the target time period output by the fault diagnosis model; Wherein, the fault diagnosis model is trained based on the fault diagnosis model training method according to any one of claims 1 to 5.

9. The fault diagnosis device according to claim 8, characterized in that: Also includes: The fault diagnosis model training device as described in claim 7.

10. A working machine, characterized in that: include: A fault diagnosis device as claimed in claim 8 or 9.

Images