Fault detection method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a fault detection method and device, electronic equipment and storage medium, relates to the technical field of artificial intelligence, in particular to the technical field of deep learning and industrial big data. The specific implementation scheme is: obtaining detection data associated with a target fault; using at least two trained detection models to detect the detection data to obtain first detection results corresponding to each detection model; performing fusion processing on the first detection results corresponding to each detection model to obtain a first target fusion result; and determining whether the target fault exists according to the first target fusion result. Thus, the accuracy and timeliness of fault detection are improved, and the labor cost is reduced.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence, in particular to the technical field of deep learning and industrial big data, and more particularly to a fault detection method and device, an electronic device and a storage medium. BACKGROUND

[0002] Currently, in order to ensure the stable operation of equipment, it is often necessary to detect whether each component in the equipment has failed. Taking the air conditioning system of an aircraft as an example, the air conditioning system can pressurize, adjust the temperature and humidity of the cabin of the aircraft, and is a key system for ensuring the comfort of the passenger cabin. At the same time, the system is susceptible to the influence of the external environment and prone to failure. Among them, the air conditioning component is the core component of refrigeration temperature control, and its performance will directly affect the operation reliability of the entire air conditioning system and the aircraft. Therefore, timely and accurate fault detection of the air conditioning component is a very necessary work. SUMMARY

[0003] The present disclosure provides a fault detection method and device, an electronic device and a storage medium.

[0004] According to an aspect of the present disclosure, a fault detection method is provided, comprising: obtaining to-be-detected data associated with a target fault; using at least two trained detection models to respectively perform fault detection on the to-be-detected data to obtain first detection results corresponding to each detection model; performing fusion processing on the first detection results corresponding to each detection model to obtain a first target fusion result; and determining whether the target fault exists according to the first target fusion result.

[0005] According to another aspect of the present disclosure, a training method of a detection model for fault detection is provided, comprising: obtaining first sample data and second sample data associated with a target fault; using the first sample data to individually train at least two detection models to obtain pre-trained at least two detection models; using the pre-trained at least two detection models to respectively perform fault detection on the second sample data to obtain second detection results corresponding to each detection model; performing fusion processing on the second detection results corresponding to each detection model to obtain a second target fusion result; and when the accuracy of the second target fusion result is less than or equal to a preset second threshold, using the first sample data to re-individually train the pre-trained at least two detection models until the accuracy of the second target fusion result is greater than the preset second threshold, thereby obtaining trained at least two detection models.

[0006] According to another aspect of the present disclosure, there is provided a fault detection apparatus, comprising: a first obtaining module configured to obtain to-be-detected data associated with a target fault; a first detection module configured to perform fault detection on the to-be-detected data using at least two trained detection models respectively, to obtain a first detection result corresponding to each of the detection models; a first fusion module configured to perform fusion processing on the first detection result corresponding to each of the detection models, to obtain a first target fusion result; and a determination module configured to determine whether the target fault exists according to the first target fusion result.

[0007] According to another aspect of the present disclosure, there is provided a training apparatus for a detection model used for fault detection, comprising: a second obtaining module configured to obtain first sample data and second sample data associated with a target fault; a training module configured to perform individual training on at least two detection models using the first sample data, to obtain the at least two pre-trained detection models; a second detection module configured to perform fault detection on the second sample data using the at least two pre-trained detection models respectively, to obtain a second detection result corresponding to each of the detection models; a second fusion module configured to perform fusion processing on the second detection result corresponding to each of the detection models, to obtain a second target fusion result; and a processing module configured to, when an accuracy of the second target fusion result is less than or equal to a preset second threshold, perform individual training on the at least two pre-trained detection models using the first sample data again, until the accuracy of the second target fusion result is greater than the preset second threshold, to obtain the at least two trained detection models.

[0008] According to another aspect of the present disclosure, there is provided an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the fault detection method according to the first aspect, or the training method for a detection model used for fault detection according to the second aspect.

[0009] According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the fault detection method according to the first aspect, or the training method for a detection model used for fault detection according to the second aspect.

[0010] According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program, which, when executed by a processor, implements the steps of the fault detection method according to the first aspect, or the steps of the training method for a detection model used for fault detection according to the second aspect.

[0011] It should be appreciated that the content described in this section is not intended to identify key or critical features of the embodiments of the present disclosure or to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0012] The accompanying drawings are used to better understand the present scheme and do not limit the present disclosure. Among them:

[0013] Figure 1 is a flowchart of a fault detection method according to a first embodiment of the present disclosure;

[0014] Figure 2 is a flowchart of a fault detection method according to a second embodiment of the present disclosure;

[0015] Figure 3 is a flowchart of a training method of a detection model for fault detection according to a third embodiment of the present disclosure;

[0016] Figure 4 is a flowchart of a training method of a detection model for fault detection according to a fourth embodiment of the present disclosure;

[0017] Figure 5 is a structural diagram of a fault detection device according to a fifth embodiment of the present disclosure;

[0018] Figure 6 is a structural diagram of a fault detection device according to a sixth embodiment of the present disclosure;

[0019] Figure 7 is a structural diagram of a training device of a detection model for fault detection according to a seventh embodiment of the present disclosure;

[0020] Figure 8 is a block diagram of an electronic device for implementing a fault detection method or a training method of a detection model for fault detection according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to help in understanding, and should be considered as merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Also, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0022] The present disclosure relates to the field of artificial intelligence technology, specifically to the fields of deep learning and industrial big data technology.

[0023] The technical field related to the scheme of the present disclosure is briefly described as follows:

[0024] AI (Artificial Intelligence) is a discipline that studies enabling computers to simulate some thinking processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.) of human beings, and has both hardware technology and software technology. Artificial intelligence hardware technology generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing, etc.; artificial intelligence software technology mainly includes computer vision, speech recognition technology, natural language processing technology, and machine learning / deep learning, big data processing technology, knowledge graph technology, etc. several major directions.

[0025] DL (Deep Learning) is to learn the internal rules and representation levels of sample data, and the information obtained in these learning processes is very helpful for the interpretation of data such as text, images and sound. The ultimate goal is to enable machines to have analysis and learning capabilities like humans, and to recognize text, images and sound data. Deep learning is a complex machine learning algorithm, and the results achieved in speech and image recognition far exceed previous related technologies. Deep learning has achieved a lot in search technology, data mining, machine learning, machine translation, natural language processing, multimedia learning, speech, recommendation and personalization technology, and other related fields. Deep learning enables machines to imitate human activities such as vision and thinking, solves many complex pattern recognition problems, and enables artificial intelligence-related technologies to make great progress.

[0026] Data processing is the collection, storage, retrieval, processing, transformation and transmission of data. According to the structure of the processing device, the working mode, and the time and space distribution mode of the data, there are different ways of data processing. Different processing methods require different hardware and software support. Each processing method has its own characteristics, and the appropriate processing method should be selected according to the actual environment of the application problem.

[0027] At present, in order to ensure the stable operation of the equipment, it is often necessary to detect whether each component in the equipment has failed.

[0028] Taking the air conditioning system of an aircraft as an example, the air conditioning system can pressurize, adjust the temperature and humidity of the cabin of the aircraft, and is a key system to ensure the comfort of the passenger cabin. At the same time, the system is susceptible to external environmental influences and prone to failure. Among them, the air conditioning component is the core component of refrigeration temperature control, and its performance will directly affect the operation reliability of the entire air conditioning system and the aircraft, therefore, it is a very necessary work to timely and accurately detect the related faults of the air conditioning component.

[0029] In the related art, the fault detection of the air conditioning assembly mainly relies on the experience formed by the maintenance manual and daily summary. Since it takes a long time to consult the maintenance manual, the problems existing in the assembly during operation cannot be found in time, and therefore, the timeliness of the fault detection is poor. In addition, since the troubleshooting and maintenance work is mainly performed manually, a large amount of labor cost is required, and it cannot be ensured that the fault can be accurately found out, and therefore, the accuracy of the fault detection is poor, and the labor cost is high.

[0030] The present disclosure provides a fault detection method, device, electronic equipment, non-transitory computer-readable storage medium and computer program product, which adopts at least two detection models for fault detection, and fuses the first detection results corresponding to each detection model to obtain a first target fusion result, and then determines whether there is a target fault according to the first target fusion result, so as to realize timely and accurate fault detection, improve the timeliness and accuracy of fault detection, and reduce the labor cost.

[0031] The fault detection method, the training method and device of the detection model for fault detection, the electronic equipment, the non-transitory computer-readable storage medium and the computer program product of the present disclosure are described below with reference to the accompanying drawings.

[0032] Firstly, the fault detection method provided by the present disclosure is described in detail. Figure 1 The fault detection method provided by the present disclosure is described in detail.

[0033] Figure 1 is a flowchart of the fault detection method according to the first embodiment of the present disclosure. It should be noted that the fault detection method provided by the present embodiment is executed by a fault detection device, which can be an electronic equipment or can be configured in an electronic equipment, so as to improve the timeliness and accuracy of fault detection and reduce the labor cost. The present embodiment takes the fault detection device configured in an electronic equipment as an example for description.

[0034] The electronic equipment can be any stationary or mobile computing device capable of data processing, such as a notebook computer, a smart phone, a wearable device, a stationary computing device such as a desktop computer, a server, or other types of computing devices, and the present disclosure does not limit the same.

[0035] As shown in Figure 1 The fault detection method can include the following steps:

[0036] Step 101, obtaining the to-be-detected data associated with the target fault.

[0037] The target fault can include fault objects, fault types, and the like. For example, the target fault can be a low flow fault of an air conditioning component, a poor heating effect fault of the air conditioning component, a poor refrigeration effect fault of a refrigeration device, and the like. The flow can refer to the flow of refrigerant.

[0038] The to-be-detected data associated with the target fault can include data collected for variables that can affect or cause the target fault. For example, when the target fault is a low flow fault of an air conditioning component in an aircraft, the flight altitude of the aircraft, the temperature of the external environment, the humidity of the external environment, and the like can affect the flow of the air conditioning component and cause the flow to be low. Therefore, the to-be-detected data associated with the low flow fault of the air conditioning component can include data collected for the flight altitude of the aircraft, the temperature of the external environment, the humidity of the external environment, and the like.

[0039] In an example embodiment, the to-be-detected data can be collected in real time by various sensors or detectors and then sent to the fault detection device, so that the fault detection device can obtain the to-be-detected data in real time.

[0040] At step 102, at least two trained detection models are used to perform fault detection on the to-be-detected data respectively to obtain first detection results corresponding to the detection models respectively.

[0041] The detection model can be a model trained in an unsupervised manner or a model trained in a supervised manner, and the disclosure does not limit this. For example, the detection model can be any model that can be trained in an unsupervised manner and can be used for fault detection, such as an IF (Isolation Forest), an Autoencoder, a LOF (Local Outlier Factor), and the like. The detection model can also be any model that can be trained in a supervised manner and can be used for fault detection, such as an LSTM (Long-Short Term Memory) network, a CNN (Convolutional Neural Network), and the like.

[0042] In an example embodiment, at least two detection models can be trained in advance respectively, and then after obtaining the to-be-detected data associated with the target fault, the to-be-detected data can be input into each detection model respectively. At least two trained detection models are used to perform fault detection on the to-be-detected data respectively, so as to obtain first detection results corresponding to the detection models respectively.

[0043] The first detection result can include a score indicating a probability of the target fault, a time when the target fault occurs, a cause of the target fault, and the like.

[0044] In step 103, the first detection results corresponding to the detection models are fused to obtain a first target fusion result.

[0045] In step 104, whether the target fault exists is determined according to the first target fusion result.

[0046] It can be understood that, because the capture effects of different detection models on features are different, the fault detection performances of the detection models are different, and thus when the detection models are used to detect faults of the to-be-detected data, the first detection results output by the detection models can be different. In the embodiment of the disclosure, the first detection results corresponding to the detection models can be fused to obtain a first target fusion result, and then whether the target fault exists is determined according to the first target fusion result.

[0047] In a possible implementation form, the first detection results corresponding to the detection models are fused to obtain the first target fusion result, and specifically, scores indicating the probability of the target fault can be weighted and averaged to obtain a first target fusion score as the first target fusion result.

[0048] For example, it is assumed that two detection models are used to detect faults of the to-be-detected data in the embodiment of the disclosure, and the two detection models output scores 0.8 and 0.6 indicating the probability of the target fault, and the weights corresponding to the two detection models are 0.5, respectively. The two scores can be weighted and averaged to obtain a first target fusion score (0.8*0.5+0.6*0.5) / (0.5+0.5)=0.7. According to the first target fusion score, whether the target fault exists can be determined.

[0049] In another possible implementation form, the first detection results corresponding to the detection models are fused to obtain the first target fusion result, and specifically, when there are at least two same first detection results in the first detection results corresponding to the at least two detection models, the same first detection result is taken as the first target fusion result.

[0050] For example, assuming that the target fault is a flow low fault of an air conditioning component, three detection models are used in the embodiments of the present disclosure to perform fault detection on the to-be-detected data. The first detection results corresponding to the three detection models are: "the air conditioning component has a flow low fault, and the fault reason is that the humidity of the external environment is too low"; "the air conditioning component has a flow low fault, and the fault reason is that the humidity of the external environment is too low"; and "the air conditioning component does not have a flow low fault". The first target fusion result can be selected as "the air conditioning component has a flow low fault, and the fault reason is that the humidity of the external environment is too low" from the three first detection results. According to the first target fusion result, it can be determined that the target fault exists.

[0051] It should be noted that the process of fusing the first detection results corresponding to the detection models to obtain the first target fusion result is only illustrative and cannot be understood as a limitation on the technical solutions of the present disclosure. In actual applications, those skilled in the art can use any other fusion method to fuse the first detection results corresponding to the detection models to obtain the first target fusion result, and the present disclosure does not limit this.

[0052] In the embodiments of the present disclosure, the first detection results corresponding to the detection models are fused to obtain the first target fusion result, and then it is determined whether the target fault exists according to the first target fusion result. The target fault can be determined according to the fault detection results of the detection models, thereby improving the accuracy of fault detection, and the generalization of the fault detection method is good. Moreover, the fault detection device can automatically perform fault detection on the to-be-detected data associated with the target fault in real time, and timely determine whether the target fault exists, thereby reducing the labor cost of fault detection and improving the timeliness of fault detection.

[0053] In summary, the fault detection method provided in the embodiments of the present disclosure acquires to-be-detected data associated with a target fault, uses at least two trained detection models to perform fault detection on the to-be-detected data to obtain first detection results corresponding to the detection models, fuses the first detection results corresponding to the detection models to obtain a first target fusion result, and determines whether the target fault exists according to the first target fusion result. The accuracy and timeliness of fault detection are improved, and the labor cost is reduced.

[0054] From the above analysis, it can be seen that in the embodiments of the present disclosure, the first detection results corresponding to the detection models can be fused to obtain the first target fusion result, and then it is determined whether the target fault exists according to the first target fusion result. Next, the fault detection method provided in the embodiments of the present disclosure will be described in detail in combination with the accompanying drawings. Figure 2The process of fusing the first detection results corresponding to each detection model and determining whether the target fault exists according to the obtained first target fusion result in the fault detection method provided by the present disclosure is further described.

[0055] Figure 2 is a flowchart of a fault detection method according to a second embodiment of the present disclosure. As shown in Figure 2 the fault detection method can include the following steps:

[0056] Step 201, obtaining detection data associated with a target fault.

[0057] The detection data can include raw data collected for at least one key variable at a plurality of first sampling time points and statistical feature data constructed according to the raw data. The statistical feature data includes feature value data of the raw data collected for each key variable at the plurality of first sampling time points.

[0058] The key variable refers to a variable that will affect or cause the target fault. For example, when the target fault is a low flow fault of an air conditioning assembly in an aircraft, the key variable can include at least one of the following variables: cruise stage of the aircraft, flight altitude of the aircraft, temperature of the external environment, humidity of the external environment, outlet temperature of the air compressor of the air conditioning system, outlet temperature of the air conditioning assembly, and bypass valve opening degree of the air conditioning system.

[0059] The raw data refers to data collected by sensors or detectors.

[0060] The feature value data can refer to maximum value, minimum value, average value, variance, standard deviation, peak difference, kurtosis factor, skewness factor, pulse factor, margin factor, and peak value factor.

[0061] The plurality of first sampling time points can be a plurality of sampling time points with a preset step length in a time period. The length of the time period, the preset step length, the number of sampling time points, and the number of key variables can be set as needed, and the present disclosure does not limit them. For example, taking one sampling time point per second as an example, the plurality of first sampling time points can be 60 sampling time points in the previous 1 minute.

[0062] Taking the target fault as a low flow fault of an air conditioning assembly in an aircraft, the key variable including flight altitude of the aircraft, temperature of the external environment, and humidity of the external environment, the feature value data including maximum value, minimum value, and average value data, and the plurality of first sampling time points as 60 sampling time points in the previous 1 minute as an example, the various data included in the detection data is described.

[0063] The to-be-detected data includes the following raw data: raw data collected at a first sampling time point x1 for the flight height of the aircraft, raw data collected at the first sampling time point x1 for the temperature of the external environment, and raw data collected at the first sampling time point x1 for the humidity of the external environment; raw data collected at a first sampling time point x2 for the flight height of the aircraft, raw data collected at the first sampling time point x2 for the temperature of the external environment, and raw data collected at the first sampling time point x2 for the humidity of the external environment; and raw data collected at first sampling time points x3 to x60 for the flight height of the aircraft, raw data collected at the first sampling time points x3 to x60 for the temperature of the external environment, and raw data collected at the first sampling time points x3 to x60 for the humidity of the external environment. In addition, the to-be-detected data also includes the following statistical feature data: maximum value, minimum value, and mean value data of the raw data collected at the 60 first sampling time points for the flight height of the aircraft; maximum value, minimum value, and mean value data of the raw data collected at the 60 first sampling time points for the temperature of the external environment; and maximum value, minimum value, and mean value data of the raw data collected at the 60 first sampling time points for the humidity of the external environment.

[0064] By collecting raw data for at least one key variable at a plurality of first sampling time points, comprehensive fault detection can be performed based on the raw data collected for a plurality of key variables within a time period. By constructing statistical feature data from the raw data and combining the raw data and the statistical feature data as to-be-detected data, the amount and types of to-be-detected data are enriched, so that more accurate fault detection can be performed based on the to-be-detected data.

[0065] In addition, when the target fault is a low-flow fault of the air conditioning assembly in the aircraft, at least one of the cruising stage of the aircraft, the flight height of the aircraft, the temperature of the external environment, the humidity of the external environment, the outlet temperature of the air compressor of the air conditioning system, the outlet temperature of the air conditioning assembly, and the opening degree of the bypass valve of the air conditioning system is used as a key variable, so that it can be comprehensively detected from each key variable dimension whether the air conditioning assembly in the aircraft has a low-flow fault, thereby laying a foundation for improving the detection accuracy of the low-flow fault of the air conditioning assembly in the aircraft, ensuring efficient and smooth operation of the aircraft, and improving passenger experience.

[0066] Step 202: using the trained at least two detection models to respectively perform fault detection on the to-be-detected data to obtain a score corresponding to each detection model indicating the probability of the existence of the target fault.

[0067] The specific implementation process and principles of step 202 can be referred to the description of the above embodiments, which will not be repeated here.

[0068] Step 203, the scores corresponding to each detection model are weighted and averaged to obtain a first target fusion score as a first target fusion result.

[0069] In an example embodiment, at least two detection models can be trained in advance respectively, and then after obtaining the to-be-detected data associated with the target fault, the to-be-detected data can be input into each detection model to perform fault detection on the to-be-detected data by using the trained at least two detection models, so as to obtain a score corresponding to each detection model indicating the probability of the existence of the target fault. Then, the scores corresponding to each detection model can be weighted and averaged to obtain a first target fusion score as a first target fusion result.

[0070] For example, assuming that two detection models are used in the embodiment of the present disclosure to perform fault detection on the to-be-detected data, and the two detection models output scores 0.8 and 0.6 indicating the probability of the existence of the target fault, and the weights corresponding to the two detection models are 0.5 respectively, then the two scores can be weighted and averaged to obtain a first target fusion score 0.7.

[0071] Step 204, comparing the first target fusion score with a preset first threshold.

[0072] Step 205, when the first target fusion score is greater than the preset first threshold, it is determined that the target fault exists.

[0073] Step 206, when the first target fusion score is less than or equal to the preset first threshold, it is determined that the target fault does not exist.

[0074] The preset first threshold can be set as needed, for example, set to 0.5, which is not limited in the present disclosure.

[0075] For example, taking the preset first threshold as 0.5, assuming that the first target fusion score is 0.7, since 0.7 is greater than 0.5, it can be determined that the target fault exists; assuming that the first target fusion score is 0.45, since 0.45 is less than 0.5, it can be determined that the target fault does not exist.

[0076] By weighting and averaging the scores corresponding to each detection model, a first target fusion score as a first target fusion result is obtained, which realizes the fusion processing of the first detection results output by each detection model, and by determining whether the target fault exists according to the size relationship between the first target fusion score and the preset first threshold, accurate fault detection is realized.

[0077] It should be noted that when it is determined that the target fault exists, a warning prompt can also be performed to timely inform the relevant staff to take measures to avoid losses caused by the target fault.

[0078] In summary, the fault detection method provided by the embodiments of the present disclosure, by obtaining the to-be-detected data associated with the target fault, using at least two trained detection models to respectively detect the to-be-detected data to obtain the score corresponding to each detection model indicating the probability of the existence of the target fault, and performing weighted average on the scores corresponding to each detection model to obtain the first target fusion score as the first target fusion result, comparing the first target fusion score with the preset first threshold, when the first target fusion score is greater than the preset first threshold, determining that the target fault exists, and when the first target fusion score is less than or equal to the preset first threshold, determining that the target fault does not exist, the accuracy and timeliness of fault detection are improved, and the labor cost is reduced.

[0079] From the above analysis, it can be known that in the embodiments of the present disclosure, at least two trained detection models can be used to detect the to-be-detected data. The process of training the at least two detection models in the above embodiments is described below.

[0080] Figure 3 is a flowchart of a training method of a detection model for fault detection according to the third embodiment of the present disclosure. It should be noted that the training method of the detection model for fault detection provided by the present embodiment is executed by a training device of the detection model for fault detection, hereinafter referred to as the training device. The training device can be an electronic device or can be configured in an electronic device to train each detection model. The present embodiment takes the training device configured in an electronic device as an example for description.

[0081] As shown in Figure 3 , the training method of the detection model for fault detection can include the following steps:

[0082] Step 301, obtaining first sample data and second sample data associated with a target fault.

[0083] The target fault can include fault objects, fault types, and other information. For example, the target fault can be a low flow fault of an air conditioning component, a poor heating effect fault of an air conditioning component, a poor refrigeration effect fault of a refrigeration device, etc. The flow rate can refer to the flow rate of refrigerant.

[0084] The first sample data or the second sample data associated with the target fault can include data collected for variables that can affect or cause the target fault. For example, when the target fault is a low flow fault of an air conditioning assembly in an aircraft, the flight altitude of the aircraft, the temperature of the external environment, the humidity of the external environment, etc. can affect the flow of the air conditioning assembly and cause the low flow fault. Therefore, the first sample data or the second sample data associated with the low flow fault of the air conditioning assembly can include data collected for the flight altitude of the aircraft, the temperature of the external environment, the humidity of the external environment, etc.

[0085] It should be noted that the first sample data or the second sample data can only include data collected for variables that can affect or cause the target fault, without marking these data as whether they are collected in the presence of the target fault, or these data can be further marked as whether they are collected in the presence of the target fault. That is, the first sample data or the second sample data can be sample data without corresponding annotations, or sample data with corresponding annotations, and the present disclosure does not limit this.

[0086] In an example embodiment, when the first sample data or the second sample data is sample data without corresponding annotations, the first sample data and the second sample data can be collected in real time by various sensors or detectors and saved to a database, so that the training device can obtain the first sample data and the second sample data from the database.

[0087] In an example embodiment, when the first sample data or the second sample data is sample data with corresponding annotations, the first sample data and the second sample data can be collected in real time by various sensors or detectors and saved to a database, and whether they are collected in the presence of the target fault is marked by artificial marking, so that the training device can obtain the first sample data and the second sample data with corresponding annotations from the database.

[0088] Step 302, training at least two detection models separately using the first sample data to obtain at least two pre-trained detection models.

[0089] The detection model can be a model that can be trained in an unsupervised manner or a model that can be trained in a supervised manner, and the disclosure does not limit this. For example, the detection model can be any model that can be trained in an unsupervised manner and can be used for fault detection, such as IF, Autoencoder, LOF, etc. The detection model can also be any model that can be trained in a supervised manner and can be used for fault detection, such as LSTM network, CNN, etc. The appropriate detection model can be selected according to the data characteristics of the first sample data and the second sample data, whether there is corresponding labeling, and the characteristics of each detection model.

[0090] In an example embodiment, the first sample data can be used to individually train each detection model, and the training method is different for different types of detection models.

[0091] For example, a detection model can be trained in a supervised manner using deep learning: at least one sample in the first sample data is input into the detection model, and the prediction information of the detection model is obtained. The prediction information is compared with the labeling information corresponding to the sample, the difference between the prediction information and the labeling information is obtained, and then the model parameters of the detection model are adjusted according to the difference to obtain an adjusted detection model. Thus, the model parameters of the detection model are repeatedly adjusted to iteratively train the detection model by continuously using the samples in the first sample data, until the accuracy of the prediction information output by the detection model meets a pre-set threshold, the training is completed, and a pre-trained detection model is obtained.

[0092] Alternatively, for example, a detection model can be trained in an unsupervised manner: each sample in the first sample data is input into the detection model to cluster each sample using the detection model, thereby distinguishing the positive sample data and the negative sample data in the first sample data, and then performing self-learning according to the positive sample data and the negative sample data to obtain a pre-trained detection model.

[0093] Step 303, using at least two pre-trained detection models, respectively, for fault detection of the second sample data to obtain the second detection result corresponding to each detection model.

[0094] In an example embodiment, the second sample data can include multiple samples, and for each sample, the sample can be input into each pre-trained detection model to use each pre-trained detection model to perform fault detection on the sample, thereby obtaining the second detection result corresponding to each pre-trained detection model.

[0095] The second detection result can include a score indicating a probability of the target fault, a time when the target fault occurs, a cause of the target fault, and the like.

[0096] At step 304, the second detection results corresponding to the detection models are fused to obtain a second target fusion result.

[0097] The specific implementation process and principle of step 304 can refer to the fusion process and principle of the first detection stage of each detection model in the above embodiments, which will not be described here.

[0098] At step 305, when the accuracy of the second target fusion result is less than or equal to a preset second threshold, the first sample data is used to retrain the at least two pre-trained detection models, until the accuracy of the second target fusion result is greater than the preset second threshold, and the trained at least two detection models are obtained.

[0099] The preset second threshold can be set as needed, and the present disclosure does not limit it.

[0100] In an exemplary embodiment, for each sample in the second sample data, after obtaining the corresponding second target fusion result, the accuracy of the second target fusion result can be determined in the following manner.

[0101] When the second sample data is labeled sample data, the second target fusion result can be compared with the label of the corresponding sample, and the accuracy of the second target fusion result can be determined according to the matching degree. For example, when the matching degree of the second target fusion result and the label of the corresponding sample is 80%, the accuracy of the second target fusion result can be determined as 80%.

[0102] When the second sample data is unlabeled sample data, the second target fusion result and the corresponding sample can be analyzed in combination with the experience of relevant experts to determine the accuracy of the second target fusion result. For example, when the second target fusion result is a score indicating the probability of the target fault, and the score is 80%, and it is assumed that the score is greater than a preset first threshold of 50%, i.e., it is determined that there is a target fault, if it is determined through the analysis of the relevant experts that the sample corresponding to the second target fusion result is collected when there is a target fault, it can be determined that the accuracy of the second target fusion result is 100%.

[0103] In the example embodiment, the accuracy of the second target fusion result corresponding to one sample can be determined first, and it is determined whether the accuracy is greater than the preset second threshold. If not, the pre-trained at least two detection models are retrained individually according to the manner of step 302. The accuracy of the second target fusion result corresponding to multiple samples can also be determined first, and it is determined whether all the multiple accuracies are greater than the preset second threshold. If only a small part is greater than the preset second threshold, the pre-trained at least two detection models are retrained individually according to the manner of step 302.

[0104] After the pre-trained at least two detection models are retrained individually, steps 303-305 can be continued to be executed until the accuracy of the second target fusion result is greater than the preset second threshold, and the trained at least two detection models are obtained. The trained at least two detection models can be used for fault detection by the above-mentioned fault detection method.

[0105] It should be noted that, in actual application, before the trained at least two detection models are used for fault detection, third sample data can also be obtained, and the trained at least two detection models are continued to be used to perform fault detection on the third sample data respectively to obtain third detection results corresponding to each detection model. The third detection results corresponding to each detection model are fused to obtain a third target fusion result, and then according to the accuracy of the third target fusion result, it is determined whether the trained at least two detection models need to be retrained individually. That is, in actual application, before the trained at least two detection models are used for fault detection, the detection effect of the at least two detection models can also be verified for multiple rounds to improve the fault detection effect of each detection model.

[0106] In summary, after the first sample data and the second sample data associated with the target fault are obtained, the first sample data is used to train the at least two detection models individually to obtain the pre-trained at least two detection models. The pre-trained at least two detection models are used to perform fault detection on the second sample data respectively to obtain second detection results corresponding to each detection model. The second detection results corresponding to each detection model are fused to obtain a second target fusion result. When the accuracy of the second target fusion result is less than or equal to the preset second threshold, the first sample data is used to retrain the pre-trained at least two detection models individually until the accuracy of the second target fusion result is greater than the preset second threshold, and the trained at least two detection models are obtained. The at least two detection models are trained to obtain the at least two detection models that can be used for the fault detection method in the above-mentioned embodiment.

[0107] The above-mentioned embodiments are described below in combination with Figure 4The training method of the detection model for fault detection in the embodiments of the present disclosure is further described.

[0108] Figure 4 is a flowchart of the training method of the detection model for fault detection according to the fourth embodiment of the present disclosure. As shown in Figure 4 , the training method of the detection model for fault detection can include the following steps:

[0109] Step 401, obtaining first sample data and second sample data associated with a target fault.

[0110] Among the first sample data or the second sample data, there can be multiple samples, and each sample can include sample raw data collected for at least one key variable at multiple second sampling time points, and sample statistical feature data constructed according to the sample raw data. Among the sample statistical feature data, there are feature value data of the sample raw data collected for each key variable at multiple second sampling time points.

[0111] Among the sample raw data, there are sample data collected by each sensor or detector.

[0112] Among the multiple second sampling time points, there can be multiple sampling time points with a preset step length in a time period, wherein the length of the time period, the preset step length, the number of sampling time points, and the number of key variables, etc. can be set as needed, and the present disclosure does not limit this. For example, taking one sampling time point per second as an example, the multiple second sampling time points can be 60 sampling time points in any 1 minute.

[0113] Next, taking the target fault as the low flow fault of the air conditioning assembly in the aircraft, the key variables including the flight height of the aircraft, the temperature of the external environment, and the humidity of the external environment, the feature value data including the maximum value, the minimum value, and the average value data, and the multiple second sampling time points as 60 sampling time points in the 1 minute before the current time as an example, the various data included in a sample in the first sample data or the second sample data is described.

[0114] Each sample can include the following sample original data: sample original data collected at a second sampling time point y1 for the flight height of the aircraft, sample original data collected for the temperature of the external environment, and sample original data collected for the humidity of the external environment; sample original data collected at a second sampling time point y2 for the flight height of the aircraft, sample original data collected for the temperature of the external environment, and sample original data collected for the humidity of the external environment; and sample original data collected at second sampling time points y3 to y60 for the flight height of the aircraft, sample original data collected for the temperature of the external environment, and sample original data collected for the humidity of the external environment, respectively. In addition, each sample also includes the following sample statistical feature data: maximum value, minimum value, and mean value data of the sample original data collected at the 60 second sampling time points for the flight height of the aircraft; maximum value, minimum value, and mean value data of the sample original data collected at the 60 second sampling time points for the temperature of the external environment; and maximum value, minimum value, and mean value data of the sample original data collected at the 60 second sampling time points for the humidity of the external environment.

[0115] By collecting sample original data for at least one key variable at multiple second sampling time points, at least two detection models for fault detection can be comprehensively trained based on sample original data collected for multiple key variables within a time period, and by constructing sample statistical feature data from sample original data and combining sample original data and sample statistical feature data as first sample data or second sample data, the data quantity and types of the first sample data and the second sample data are enriched, so that based on the first sample data and the second sample data, each detection model can be trained to have better effect.

[0116] Step 402, the first sample data is used to individually train at least two detection models to obtain at least two pre-trained detection models.

[0117] Step 403, the at least two pre-trained detection models are used to respectively perform fault detection on the second sample data to obtain second detection results corresponding to each detection model.

[0118] Step 404, the second detection results corresponding to each detection model are fused to obtain a second target fusion result.

[0119] The specific implementation process and principles of steps 402-404 can be referred to the description of the above embodiments, which will not be repeated here.

[0120] At step 405, the accuracy of the second target fusion result is determined, and it is determined whether the accuracy is greater than a preset second threshold. If yes, step 406 is executed, otherwise, step 402 is returned to be executed until the accuracy is greater than the preset second threshold.

[0121] At step 406, the at least two detection models trained are obtained.

[0122] In the exemplary embodiment, the accuracy of the second target fusion result corresponding to a sample can be determined first, and it is determined whether the accuracy is greater than a preset second threshold. If not, step 402 is returned immediately to retrain the at least two detection models pre-trained. Alternatively, the accuracy of the second target fusion result corresponding to multiple samples can be determined first, and it is determined whether all the accuracies are greater than a preset second threshold. If only a small part of the accuracies are greater than the preset second threshold, step 402 can be returned to retrain the at least two detection models pre-trained.

[0123] After the at least two detection models pre-trained are retrained, step 403 and the steps thereafter can be continued to be executed until the accuracy of the second target fusion result is greater than the preset second threshold, and the at least two detection models trained are obtained. The at least two detection models trained can be used for fault detection by the above fault detection method.

[0124] In summary, after the first sample data and the second sample data associated with the target fault are obtained, the first sample data is used to train the at least two detection models to obtain the at least two detection models pre-trained. The at least two detection models pre-trained are used to perform fault detection on the second sample data respectively to obtain the second detection results corresponding to the detection models. The second detection results corresponding to the detection models are fused to obtain the second target fusion result. The accuracy of the second target fusion result is determined, and it is determined whether the accuracy is greater than a second threshold. If the accuracy of the second target fusion result is less than or equal to the preset second threshold, the first sample data is used to retrain the at least two detection models pre-trained. Until the accuracy of the second target fusion result is greater than the preset second threshold, the at least two detection models trained are obtained. The at least two detection models are trained to obtain the at least two detection models that can be used in the fault detection method in the above embodiment.

[0125] The following will be described in combination with Figure 5 The fault detection device provided by the present disclosure is described.

[0126] Figure 5 FIG. 5 is a structural schematic diagram of a fault detection device according to the fifth embodiment of the present disclosure.

[0127] As Figure 5As shown, the fault detection apparatus 500 provided by the present disclosure includes a first acquisition module 501, a first detection module 502, a first fusion module 503, and a determination module 504.

[0128] The first acquisition module 501 is configured to acquire detection data associated with a target fault.

[0129] The first detection module 502 is configured to perform fault detection on the detection data by using at least two trained detection models respectively, to obtain first detection results corresponding to the detection models.

[0130] The first fusion module 503 is configured to perform fusion processing on the first detection results corresponding to the detection models, to obtain a first target fusion result.

[0131] The determination module 504 is configured to determine whether the target fault exists according to the first target fusion result.

[0132] It should be noted that the fault detection apparatus provided by the present embodiment can perform the fault detection method of the foregoing embodiments. The fault detection apparatus can be an electronic device or can be configured in an electronic device, so as to improve the timeliness and accuracy of fault detection.

[0133] The electronic device can be any stationary or mobile computing device capable of data processing, such as a notebook computer, a smart phone, a wearable device, or a stationary computing device such as a desktop computer, a server, or other types of computing devices, without limitation.

[0134] It should be noted that the foregoing description of the embodiments of the fault detection method also applies to the fault detection apparatus provided by the present disclosure, which will not be described herein again.

[0135] The fault detection apparatus provided by the present embodiment improves the accuracy and timeliness of fault detection, and reduces labor costs, by acquiring detection data associated with a target fault, performing fault detection on the detection data by using at least two trained detection models respectively, to obtain first detection results corresponding to the detection models, performing fusion processing on the first detection results corresponding to the detection models, to obtain a first target fusion result, and determining whether the target fault exists according to the first target fusion result.

[0136] The fault detection apparatus provided by the present disclosure will be described below in combination with Figure 6

[0137] Figure 6 FIG. 1 is a structural schematic diagram of a fault detection apparatus according to a sixth embodiment of the present disclosure.

[0138] As shown in FIG. 1, the fault detection apparatus 500 includes a first acquisition module 501, a first detection module 502, a first fusion module 503, and a determination module 504. Figure 6 ​As shown, the fault detection apparatus 600 can specifically include a first acquisition module 601, a first detection module 602, a first fusion module 603, and a determination module 604. Among them, Figure 6 The first acquisition module 601, the first detection module 602, the first fusion module 603, and the determination module 604 in the first acquisition module 601 have the same functions and structures. Figure 5 The first acquisition module 501, the first detection module 502, the first fusion module 503, and the determination module 504 have the same functions and structures.

[0139] In an example embodiment, the first detection result is a score indicating the size of the probability of the existence of the target fault.

[0140] The first fusion module 603 includes:

[0141] The fusion unit 6031 is configured to perform weighted average on the scores corresponding to each detection model to obtain a first target fusion score as a first target fusion result.

[0142] In an example embodiment, the determination module 604 includes:

[0143] The comparison unit 6041 is configured to compare the first target fusion score with a preset first threshold value.

[0144] The first determination unit 6042 is configured to determine that the target fault exists when the first target fusion score is greater than the preset first threshold value.

[0145] The second determination unit 6043 is configured to determine that the target fault does not exist when the first target fusion score is less than or equal to the preset first threshold value.

[0146] In an example embodiment, the to-be-detected data includes raw data collected for at least one key variable at a plurality of first sampling time points and statistical feature data constructed according to the raw data.

[0147] Among them, the statistical feature data includes feature value data of the raw data collected for each key variable at a plurality of first sampling time points.

[0148] In an example embodiment, the target fault is a flow rate fault of an air conditioning assembly in an aircraft.

[0149] The key variable includes at least one of the following variables: a cruise phase of the aircraft, a flight height of the aircraft, a temperature of an external environment, a humidity of the external environment, an outlet temperature of a compressor of the air conditioning system, an outlet temperature of the air conditioning assembly, and a bypass valve opening degree of the air conditioning system.

[0150] It should be noted that the foregoing description of the embodiments of the fault detection method also applies to the fault detection device provided by the present disclosure, which will not be described here.

[0151] The fault detection device provided by the embodiments of the present disclosure acquires the to-be-detected data associated with the target fault, uses the at least two trained detection models to perform fault detection on the to-be-detected data respectively, obtains the first detection result corresponding to each detection model, performs fusion processing on the first detection result corresponding to each detection model to obtain a first target fusion result, and determines whether the target fault exists according to the first target fusion result, thereby improving the accuracy and effectiveness of fault detection and reducing the labor cost.

[0152] The following will be described in combination with Figure 7 The training device for the detection model for fault detection provided by the present disclosure is described.

[0153] Figure 7 FIG. 7 is a structural schematic diagram of the training device for the detection model for fault detection according to the seventh embodiment of the present disclosure.

[0154] As shown in Figure 7 The training device for the detection model for fault detection 700 can specifically include:

[0155] The second acquisition module 701 is configured to acquire first sample data and second sample data associated with a target fault.

[0156] The training module 702 is configured to use the first sample data to individually train at least two detection models to obtain at least two pre-trained detection models.

[0157] The second detection module 703 is configured to use the at least two pre-trained detection models to respectively perform fault detection on the second sample data to obtain a second detection result corresponding to each detection model.

[0158] The second fusion module 704 is configured to perform fusion processing on the second detection result corresponding to each detection model to obtain a second target fusion result.

[0159] The processing module 705 is configured to, when the accuracy of the second target fusion result is less than or equal to a preset second threshold, use the first sample data to individually retrain the at least two pre-trained detection models until the accuracy of the second target fusion result is greater than the preset second threshold, and obtain the at least two trained detection models.

[0160] In exemplary embodiments, the first sample data or the second sample data includes sample raw data collected for at least one key variable at a plurality of second sampling time points and sample statistical feature data constructed according to the sample raw data.

[0161] The sample statistical feature data includes, for each key variable, feature value data of sample original data collected at a plurality of second sampling time points.

[0162] It should be noted that the foregoing description of the embodiments of the method for training the detection model for fault detection also applies to the training device for the detection model for fault detection provided by the present disclosure, which will not be described here.

[0163] The training device for the detection model for fault detection provided by the embodiments of the present disclosure obtains first sample data and second sample data associated with a target fault, separately trains at least two detection models using the first sample data to obtain at least two pre-trained detection models, respectively performs fault detection on the second sample data using the at least two pre-trained detection models to obtain second detection results corresponding to each detection model, performs fusion processing on the second detection results corresponding to each detection model to obtain a second target fusion result, re-trains the at least two pre-trained detection models using the first sample data when the accuracy of the second target fusion result is less than or equal to a preset second threshold, until the accuracy of the second target fusion result is greater than the preset second threshold, and obtains the at least two trained detection models, thereby realizing training of the at least two detection models to obtain the at least two detection models that can be used in the fault detection method in the foregoing embodiments.

[0164] According to the embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium and a computer program product.

[0165] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit implementations of the present disclosure described and / or claimed in this document.

[0166] As Figure 8As shown, the device 800 includes a computing unit 801 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0167] A plurality of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0168] The computing unit 801 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 801 performs various methods and processes described above, such as a fault detection method or a training method of a detection model for fault detection. For example, in some embodiments, the fault detection method or the training method of the detection model for fault detection can be implemented as a computer software program that is tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the fault detection method or the training method of the detection model for fault detection described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured to perform the fault detection method or the training method of the detection model for fault detection by any other appropriate means, such as by means of firmware.

[0169] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a load programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0170] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, or entirely on a remote machine or server.

[0171] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0172] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0173] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0174] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS service ("Virtual Private Server", or simply "VPS"). The server can also be a server of a distributed system, or a server combined with a blockchain.

[0175] It should be understood that various forms of flow shown above can be used, with steps reordered, added, or removed. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, without limitation herein, as long as the desired results of the technical solutions of the present disclosure can be achieved.

[0176] The above detailed description does not limit the scope of the disclosure. Various modifications, combinations, sub-combinations and alternatives can be made to the detailed description. Any modification, equivalent replacement and improvement etc. made within the spirit and principle of the disclosure shall be included in the scope of the disclosure.

Claims

1. A method for training a detection model for fault detection, comprising: obtaining first sample data and second sample data associated with a target fault, wherein the first sample data or the second sample data comprises sample raw data collected for at least one key variable at a plurality of second sampling time points, and sample statistical feature data constructed from the sample raw data, wherein the key variable refers to a variable that has an impact on or causes the target fault; individually training at least two detection models using the first sample data to obtain the at least two pre-trained detection models; performing fault detection on the second sample data using the at least two pre-trained detection models respectively to obtain second detection results corresponding to the detection models respectively; performing fusion processing on the second detection results corresponding to the detection models to obtain a second target fusion result; when the accuracy of the second target fusion result is less than or equal to a preset second threshold, re-training the at least two pre-trained detection models using the first sample data until the accuracy of the second target fusion result is greater than the preset second threshold to obtain the at least two trained detection models.

2. The method of claim 1, wherein, The sample statistical feature data comprises feature value data of the sample raw data collected for each key variable at the plurality of second sampling time points. 3.A method for fault detection, comprising: obtaining to-be-detected data associated with a target fault; performing fault detection on the to-be-detected data using at least two detection models trained by the method of any one of claims 1-2 to obtain first detection results corresponding to the detection models respectively; performing fusion processing on the first detection results corresponding to the detection models to obtain a first target fusion result; determining whether the target fault exists according to the first target fusion result.

4. The method of claim 3, wherein, The first detection result is a score indicating the probability of the existence of the target fault; The fusion processing on the first detection results corresponding to the detection models to obtain the first target fusion result comprises: performing weighted averaging on the scores corresponding to the detection models to obtain a first target fusion score as the first target fusion result.

5. The method of claim 4, wherein, The determination of whether the target fault exists according to the first target fusion result comprises: comparing the first target fusion score with a preset first threshold; when the first target fusion score is greater than the preset first threshold, determining that the target fault exists; when the first target fusion score is less than or equal to the preset first threshold, determining that the target fault does not exist.

6. The method according to any one of claims 3-5, wherein, The to-be-detected data comprises raw data collected for at least one key variable at a plurality of first sampling time points, and statistical feature data constructed from the raw data; wherein the statistical feature data comprises feature value data of the raw data collected for each key variable at the plurality of first sampling time points.

7. The method of claim 6, wherein, The target fault is a flow low fault of an air conditioning component in an aircraft. The key variables include at least one of a cruise phase of the aircraft, a flight altitude of the aircraft, a temperature of an external environment, a humidity of the external environment, an outlet temperature of a compressor of an air conditioning system, an outlet temperature of the air conditioning component, and a bypass valve opening degree of the air conditioning system.

8. A fault detection apparatus, comprising: a first obtaining module configured to obtain to-be-detected data associated with a target fault; a first detection module configured to perform fault detection on the to-be-detected data by using at least two detection models trained by the method in any one of claims 1-2, to obtain a first detection result corresponding to each of the detection models; a first fusion module configured to perform fusion processing on the first detection results corresponding to each of the detection models, to obtain a first target fusion result; a determination module configured to determine, according to the first target fusion result, whether the target fault exists.

9. The apparatus of claim 8, wherein, The first detection result is a score indicating a probability of existence of the target fault. The first fusion module comprises: a fusion unit configured to perform weighted averaging on the scores corresponding to each of the detection models, to obtain a first target fusion score as the first target fusion result.

10. The apparatus of claim 9, wherein, The determination module comprises: a comparison unit configured to compare the first target fusion score with a preset first threshold value; a first determination unit configured to determine that the target fault exists when the first target fusion score is greater than the preset first threshold value; a second determination unit configured to determine that the target fault does not exist when the first target fusion score is less than or equal to the preset first threshold value.

11. The apparatus of any of claims 8-10, wherein, The to-be-detected data comprises original data collected for at least one key variable at a plurality of first sampling time points, and statistical feature data constructed according to the original data. The statistical feature data comprises feature value data of the original data collected for each of the key variables at the plurality of first sampling time points.

12. The apparatus of claim 11, wherein, The target fault is a flow low fault of an air conditioning component in an aircraft. The key variables include at least one of a cruise phase of the aircraft, a flight altitude of the aircraft, a temperature of an external environment, a humidity of the external environment, an outlet temperature of a compressor of an air conditioning system, an outlet temperature of the air conditioning component, and a bypass valve opening degree of the air conditioning system.

13. A training apparatus of a detection model for fault detection, comprising: a second obtaining module configured to obtain first sample data and second sample data associated with a target fault, wherein the first sample data or the second sample data comprises sample original data collected for at least one key variable at a plurality of second sampling time points, and sample statistical feature data constructed according to the sample original data, wherein the key variable refers to a variable that will have an impact on the target fault or cause the target fault to occur. The training module is configured to train at least two detection models separately using the first sample data to obtain the at least two pre-trained detection models; The second detection module is configured to perform fault detection on the second sample data respectively using the at least two pre-trained detection models to obtain second detection results corresponding to the detection models respectively; The second fusion module is configured to perform fusion processing on the second detection results corresponding to the detection models to obtain a second target fusion result; The processing module is configured to re-train the at least two pre-trained detection models separately using the first sample data when the accuracy of the second target fusion result is less than or equal to a preset second threshold, until the accuracy of the second target fusion result is greater than the preset second threshold, to obtain the at least two trained detection models.

14. The apparatus of claim 13, wherein, The sample statistical feature data includes, for each key variable, feature value data of sample original data collected at the plurality of second sampling time points.

15. An electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5, or perform the method of any one of claims 6-7.

16. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method of any one of claims 1-2, or perform the method of any one of claims 3-7.

17. A computer program product comprising a computer program, which, when executed by a processor, implements the steps of the method of any one of claims 1-2, or implements the steps of the method of any one of claims 3-7.

Images