CN118035730A - Equipment fault detection method, electronic equipment and storage medium - Google Patents
Equipment fault detection method, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN118035730A CN118035730A CN202410427341.0A CN202410427341A CN118035730A CN 118035730 A CN118035730 A CN 118035730A CN 202410427341 A CN202410427341 A CN 202410427341A CN 118035730 A CN118035730 A CN 118035730A
- Authority
- CN
- China
- Prior art keywords
- network
- equipment
- data
- vector
- generation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 105
- 238000003860 storage Methods 0.000 title claims abstract description 9
- 239000013598 vector Substances 0.000 claims abstract description 133
- 238000012549 training Methods 0.000 claims abstract description 128
- 238000000034 method Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 15
- 238000012795 verification Methods 0.000 claims description 47
- 238000000605 extraction Methods 0.000 claims description 29
- 238000013507 mapping Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 8
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 238000010606 normalization Methods 0.000 description 6
- 230000001965 increasing effect Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000001364 causal effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000010339 dilation Effects 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The application belongs to the field of artificial intelligence, and provides an equipment fault detection method, electronic equipment and a storage medium. The method comprises the steps of carrying out reconstruction processing on equipment training data based on a first generation network to obtain equipment reconstruction data, determining a network loss value based on a first distance between the equipment reconstruction data and the equipment training data and a second distance between a reconstruction vector and a standard vector, adjusting the first generation network based on the network loss value and the second generation network to obtain a sample generation model, expanding a fault sample in the data sample based on the sample generation model to obtain expanded data, training a preset detection network based on the data sample and the expanded data to obtain a fault detection model, determining parameter characteristics according to the equipment type of equipment to be detected, collecting parameter information of the equipment to be detected based on the parameter characteristics, and inputting the equipment type and the parameter information into the fault detection model to obtain a detection result. The method can realize effective detection of the equipment.
Description
Technical Field
The application belongs to the technical field of artificial intelligence, and particularly relates to an equipment fault detection method, electronic equipment and a storage medium.
Background
In an actual industrial production environment, since the time period of the equipment in a normal stage is far longer than the time period of the equipment in an abnormal stage, when the diagnosis result of the equipment is uncertain, the diagnosis result of the equipment is often prone to be classified into the normal result in the conventional manner, however, the conventional manner cannot effectively diagnose the faults of the equipment in the industrial environment.
Disclosure of Invention
The application provides an equipment fault detection method, electronic equipment and a storage medium, which are used for solving the technical problem that equipment cannot be effectively detected.
An embodiment of the present application provides a method for detecting a device failure, where the method includes: reconstructing the acquired device training data based on a first pre-constructed generation network to obtain device reconstruction data; determining a network loss value of the first generation network based on a first distance between the equipment reconstruction data and the equipment training data and a second distance between a reconstruction vector corresponding to the equipment reconstruction data and a standard vector corresponding to the equipment training data; based on the network loss value and a second generation network corresponding to the first generation network, adjusting the first generation network to obtain a sample generation model; expanding a fault sample in the acquired data sample based on the sample generation model to obtain expanded data; training a preset detection network based on the data sample and the expansion data to obtain a fault detection model; determining parameter characteristics according to the equipment type of equipment to be tested, and acquiring parameter information of the equipment to be tested based on the parameter characteristics; and inputting the equipment type and the parameter information into the fault detection model to obtain a detection result of the equipment to be detected.
According to an embodiment of the present application, the first generating network includes an initial encoder network and an initial decoder network, the initial encoder network includes a vector mapping layer, a first parameter importance identifying layer, a first encoding network layer, a second parameter importance identifying layer and a second encoding network layer, the reconstructing processing is performed on the acquired device training data based on the pre-constructed first generating network, and obtaining device reconstruction data includes: mapping the equipment training data based on the vector mapping layer to obtain first coding information of the equipment training data; identifying the first coding information based on the first parameter importance identification layer, determining a first parameter weight of the equipment training data, and coding the first coding information based on the first coding network layer and the first parameter weight to obtain second coding information of the equipment training data; identifying the second coding information based on the second parameter importance identification layer, determining a second parameter weight of the equipment training data, and coding the second coding information based on the second coding network layer and the second parameter weight to obtain third coding information of the initial encoder network to the equipment training data; and mapping the third coding information and random noise based on the initial decoder network to obtain the equipment reconstruction data.
According to an embodiment of the present application, the encoding the first encoded information based on the first encoded network layer and the first parameter weight, to obtain second encoded information of the device training data includes: converting the first coding information into a first input vector of the first coding network layer based on the first parameter weights; coding the first input vector according to a first weight matrix and a first bias vector of the first coding network layer to obtain characteristic information; and activating the characteristic information according to the activation function of the first coding network layer to obtain the second coding information.
According to an embodiment of the present application, the mapping the third encoded information and random noise based on the initial decoder network, to obtain the device reconstruction data includes: calculating the element sum of each vector element in the third coding information and the vector element at the corresponding position in the random noise to obtain a second input vector of the initial decoder network; determining an output vector according to a second weight matrix, a second bias vector and the second input vector in the initial decoder network; and determining information corresponding to the output vector as the device reconstruction data according to an element mapping table in the initial decoder network.
According to an embodiment of the present application, the determining the network loss value of the first generating network based on the first distance between the device reconstruction data and the device training data, and the second distance between the reconstruction vector corresponding to the device reconstruction data and the standard vector corresponding to the device training data includes: vector encoding is carried out on the equipment reconstruction data to obtain the reconstruction vector, and vector encoding is carried out on the equipment training data to obtain a training vector; calculating the distance between the reconstruction vector and the training vector as the first distance; determining a device model corresponding to the device training data, and determining a standard vector corresponding to the device training data according to the device model and a sample type corresponding to the device training data; and calculating the distance between the reconstruction vector and the standard vector as the second distance, and carrying out weighted sum operation on the first distance and the second distance to obtain the network loss value.
According to an embodiment of the present application, the adjusting the first generation network based on the network loss value and a second generation network corresponding to the first generation network, to obtain a sample generation model includes: based on the network loss value, adjusting network parameters of the first generation network to obtain an adjusted first generation network; inputting the training data of the equipment into the adjusted first generation network to obtain first verification data; reconstructing the first verification data based on a reconstruction network layer in the second generation network to obtain second verification data; identifying the second verification data based on a label identification layer in the second generation network to obtain a data label of the first verification data; counting the number of first verification data of which the data tag is a preset tag as a first number, and counting the total number of the first verification data as a second number; if the ratio of the first quantity to the second quantity is greater than or equal to a preset ratio threshold, determining the adjusted first generation network as the sample generation model; or readjusting network parameters of the first generation network if the ratio of the first number to the second number is smaller than the preset ratio threshold.
According to an embodiment of the present application, the adjusting the network parameters of the first generating network based on the network loss value, to obtain the adjusted first generating network includes: calculating the ratio of the network loss value to the set value to obtain an adjustment ratio; adjusting the network parameters based on the adjustment proportion to obtain adjusted parameter values; and determining the adjusted first generation network based on the network structure of the first generation network and the adjusted parameter value.
According to an embodiment of the present application, the fault detection model includes a feature extraction network and a fault discrimination network, and the inputting the device type and the parameter information into the fault detection model includes: locating a target extraction network from the feature extraction networks based on the device type; performing feature extraction on the parameter information based on the target extraction network to obtain target features; inputting the target features into the fault discrimination network to obtain discrimination results of a plurality of time convolution layers in the fault discrimination network on the target features; and determining the detection result according to the discrimination results of the plurality of time convolution layers on the target features.
A second aspect of an embodiment of the present application provides an apparatus for detecting a device failure, including: the reconstruction unit is used for carrying out reconstruction processing on the acquired equipment training data based on a first pre-constructed generation network to obtain equipment reconstruction data; a determining unit, configured to determine a network loss value of the first generation network based on a first distance between the device reconfiguration data and the device training data, and a second distance between a reconfiguration vector corresponding to the device reconfiguration data and a standard vector corresponding to the device training data; the adjusting unit is used for adjusting the first generation network based on the network loss value and a second generation network corresponding to the first generation network to obtain a sample generation model; the expansion unit is used for expanding a fault sample in the acquired data samples based on the sample generation model to obtain expanded data; the training unit is used for training a preset detection network based on the data sample and the expansion data to obtain a fault detection model; the acquisition unit is used for determining parameter characteristics according to the equipment type of the equipment to be detected and acquiring parameter information of the equipment to be detected based on the parameter characteristics; and the input unit is used for inputting the equipment type and the parameter information into the fault detection model to obtain a detection result of the equipment to be detected.
A third aspect of an embodiment of the present application provides an electronic device, including: a memory storing computer readable instructions; and a processor executing computer readable instructions stored in the memory to implement the device failure detection method.
A fourth aspect of the embodiments of the present application provides a computer-readable storage medium having stored therein computer-readable instructions that are executed by a processor in an electronic device to implement the device failure detection method.
According to the technical scheme, the network loss value is determined by combining the first distance and the second distance, and the network loss value can be determined in a plurality of modes, so that the accuracy of the network loss value is improved, and the training effect of the sample generation model is improved; according to the embodiment of the application, the first generation network is verified and adjusted by combining the second generation network, and because the second generation network corresponds to the first generation network, the influence of the network structure on the verification of the first generation network can be eliminated, so that the training effect of the sample generation model can be further improved, the generation rationality and the authenticity of the expansion data are improved, and meanwhile, the second generation network is not required to be constructed when the first generation network is optimized, so that the training efficiency of the sample generation model can be improved; according to the embodiment of the application, the fault samples are expanded through the sample generation model obtained through training, and normal samples in the data samples are not required to be expanded, so that the number of the fault samples can be increased, the problem that the fault detection model cannot be trained to an optimal state due to the fact that the number of the fault samples is too low is avoided, and the detection accuracy of the fault detection model is improved; the embodiment of the application combines the equipment type of the equipment to be detected, can screen out the parameter information related to the detection result for analysis, and can improve the accuracy of the detection result and realize the effective detection of the equipment because the interference of irrelevant data is eliminated and the detection result is determined by using the fault detection model with high detection accuracy.
Drawings
Fig. 1 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Fig. 2 is a flowchart of an apparatus fault detection method provided in an embodiment of the present application.
Fig. 3 is a schematic diagram of a network structure of a first generation network according to an embodiment of the present application.
Fig. 4 is a schematic diagram of a network structure of a fault detection model according to an embodiment of the present application.
Fig. 5 is a functional block diagram of an apparatus for detecting a device failure according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described in detail with reference to the accompanying drawings and specific embodiments.
In the present application, "at least one" means one or more, and "a plurality" means two or more. "and/or", describes an association relationship of an association object, and the representation may have three relationships, for example, a and/or B may represent: a alone, a and B together, and B alone, wherein a, B may be singular or plural. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order.
In embodiments of the application, words such as "exemplary" or "such as" are used to mean serving as an example, instance, or illustration. Any embodiment or design described herein as "exemplary" or "e.g." in an embodiment should not be taken as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "such as" is intended to present related concepts in a concrete fashion. The following embodiments and features of the embodiments may be combined with each other without conflict.
Fig. 1 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
In an embodiment of the present application, the device fault detection method is applied to one or more electronic devices 1, and the electronic devices 1 include, but are not limited to, a memory 12, a processor 13, and computer readable instructions, such as a device fault detection program, stored in the memory 12 and executable on the processor 13.
It will be appreciated by those skilled in the art that the schematic diagram is merely an example of the electronic device 1 and does not constitute a limitation of the electronic device 1, and may include more or fewer components than shown, or may combine certain components, or different components, e.g. the electronic device 1 may also include input-output devices, network access devices, buses, etc.
The Processor 13 may be a central processing unit (Central Processing Unit, CPU), other general purpose Processor, digital signal Processor (DIGITAL SIGNAL Processor, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc. The general purpose processor may be a microprocessor or a processor, or any conventional processor, etc., and the processor 13 is an operation core and a control center of the electronic device 1, connects various parts of the entire electronic device 1 using various interfaces and lines, and executes an operating system of the electronic device 1 and various applications, program codes, etc. installed.
The memory 12 may be an external memory and/or an internal memory of the electronic device 1. Further, the memory 12 may be a physical memory, such as a memory bank, a TF card (Trans-FLASH CARD), or the like.
In connection with fig. 2, the memory 12 in the electronic device 1 stores computer readable instructions, and the processor 13 may execute the computer readable instructions stored in the memory 12 to implement a plurality of processes as shown in fig. 2 to implement a device failure detection method.
Fig. 2 is a flowchart of an apparatus fault detection method according to an embodiment of the present application. The order of the steps in the flowchart may be changed and some steps may be omitted according to various needs. The equipment fault detection method comprises the following steps.
S201, reconstructing the acquired device training data based on a first pre-constructed generation network to obtain device reconstruction data.
In at least one embodiment of the present application, please refer to a network structure diagram of the first generation network in fig. 3, wherein the first generation network includes an initial encoder network and an initial decoder network as shown in fig. 3. The initial encoder network includes a vector mapping layer, a first parameter importance identification layer, a first encoding network layer, a second parameter importance identification layer, and a second encoding network layer. The vector mapping layer is used for directly converting the device training data into vector representations, the first parameter importance identification layer is used for identifying the importance of each parameter in the device training data, the first coding network layer is used for enhancing the vector representations of important parameters in the device training data, the second parameter importance identification layer is used for identifying the importance of each parameter in the second coding information, and the second coding network layer is used for enhancing the vector representations of important parameters in the second coding information. The initial encoder network is configured to convert the device training data into third encoded information in which a vector representation of the significant parameters in the device training data is enhanced. The initial decoder network is used for mapping the third encoded information output by the initial encoder into device reconstruction data.
The equipment training data includes operating parameter information of the rotating machinery equipment, the mechanical and electrical equipment, and other automation equipment, wherein the operating parameter information may include, but is not limited to: temperature, rotational speed, current, torque, power, etc. The device reconfiguration data may include operating parameter information similar to the device training data, for example, the device training data may include a temperature of 30 degrees celsius for the rotating machine device, and the device reconfiguration data corresponding to the device training data may be 29.5 degrees celsius.
In at least one embodiment of the present application, the electronic device performs reconstruction processing on acquired device training data based on a first generation network constructed in advance, and obtaining device reconstruction data includes: the electronic device maps the device training data based on the vector mapping layer to obtain first coding information of the device training data. The electronic equipment identifies the first coding information based on the first parameter importance identification layer, determines first parameter weights of the equipment training data, and codes the first coding information based on the first coding network layer and the first parameter weights to obtain second coding information of the equipment training data. The electronic equipment identifies the second coding information based on the second parameter importance identification layer, determines second parameter weight of the equipment training data, and codes the second coding information based on the second coding network layer and the second parameter weight to obtain third coding information of the initial coder network to the equipment training data. The electronic device maps the third encoded information and the random noise based on the initial decoder network to obtain device reconstruction data.
Wherein the first encoded information is a directly encoded representation of the device training data by the first generation network. The first parameter weight and the second parameter weight respectively comprise importance of each parameter in the device training data, for example, the device training data comprises parameter information of temperature, rotating speed, current, torque and power, and then the first parameter weight and the second parameter weight of the device training data can be (0.1,0.25,0.2,0.25,0.2) and (0.1,0.35,0.1,0.35,0.1). The second encoded information enhances the vector representation of the important parameter in the first encoded information and the third encoded information enhances the vector representation of the important parameter in the second encoded information. The random noise may be a random vector representation.
According to the embodiment, the device training data is directly encoded through the vector mapping layer to obtain the first encoded information, the first encoded information is further identified based on the first parameter important identification layer, the first parameter weight in the device training data can be determined, the first encoded information is recoded by combining the first parameter weight and the first encoded network layer, the vector representation of important parameters is enhanced in the second encoded information output by the first encoded network layer, the second encoded information is recoded by further combining the determined second parameter weight and the second encoded network layer, the vector representation of the important parameters is further enhanced in the third encoded information output by the second encoded network layer, the third encoded information is decoded by combining random noise, and the rationality of the device reconstruction data is improved.
Specifically, the first parameter importance identification layer comprises a normalization function, a global average pooling function, a full connection layer and an activation function. The electronic equipment performs normalization processing on the first coding information by using a normalization function to obtain a normalization code, and performs dimension reduction processing on the normalization code by using a global average pooling function to obtain a low-dimension code. And the electronic equipment performs full-connection processing on the low-dimensional code by using the full-connection layer, and performs activation processing on information output by the full-connection layer by using an activation function to obtain a first parameter weight. Wherein the elements in the low-dimensional coding are the average value of each dimension in the normalized coding.
In other embodiments, the determination of the second parameter weights is similar to the determination of the first parameter weights, which will not be repeated in the present application.
Specifically, the electronic device encodes the first encoded information based on the first encoded network layer and the first parameter weight, and obtaining the second encoded information of the device training data includes: the electronic equipment converts the first coding information into a first input vector of a first coding network layer based on the first parameter weight, and codes the first input vector according to a first weight matrix and a first bias vector of the first coding network layer to obtain the characteristic information. And the electronic equipment performs activation processing on the characteristic information according to the activation function of the first coding network layer to obtain second coding information. Wherein the element in the first input vector is a product of each element in the first encoded information and an element at a corresponding position in the first parameter weight, e.g., the first encoded information is (0,1,1,0,1), the first parameter weight is (0.1,0.25,0.2,0.25,0.2), then the first input vector may be (0,0.25,0.2,0,0.2). The electronic equipment calculates the product of the first input vector and the first weight matrix, and calculates the sum of each element in the product and the element at the corresponding position in the first offset vector to obtain the characteristic information. According to the embodiment, the first parameter weight is utilized, the vector representation of important parameters in the first input vector can be enhanced, and further the first input vector is analyzed through the first coding network layer, so that deep feature expression can be extracted.
In other embodiments, the determination of the third encoded information is similar to the determination of the second encoded information, and the present application will not be repeated.
Specifically, the electronic device maps the third encoded information and the random noise based on the initial decoder network, and obtaining device reconstruction data includes: the electronic device calculates the element sum of each vector element in the third coding information and the vector element at the corresponding position in the random noise to obtain a second input vector of the initial decoder network, and determines an output vector according to the second weight matrix, the second bias vector and the second input vector in the initial decoder network. The electronic device determines information corresponding to the output vector as device reconstruction data according to an element mapping table in the initial decoder network. The embodiment decodes the third encoded information by combining random noise, and can improve the anti-interference performance of the device reconstruction data.
S202, determining a network loss value of the first generation network based on a first distance between the equipment reconstruction data and the equipment training data and a second distance between a reconstruction vector corresponding to the equipment reconstruction data and a standard vector corresponding to the equipment training data.
In at least one embodiment of the application, the first distance represents a gap between the device reconstruction data and the device training data, and the second distance represents a gap between the device reconstruction data and standard data of the same device model and the same sample type.
In at least one embodiment of the present application, determining, by the electronic device, a network loss value of the first generation network based on a first distance between the device reconstruction data and the device training data, and a second distance between a reconstruction vector corresponding to the device reconstruction data and a standard vector corresponding to the device training data includes: the electronic equipment performs vector coding on the equipment reconstruction data to obtain a reconstruction vector, performs vector coding on the equipment training data to obtain a training vector, and calculates the distance between the reconstruction vector and the training vector as a first distance. The electronic equipment determines the equipment model corresponding to the equipment training data, determines a standard vector corresponding to the equipment training data according to the equipment model and the sample type corresponding to the equipment training data, calculates the distance between the reconstructed vector and the standard vector as a second distance, and performs weighted sum operation on the first distance and the second distance to obtain a network loss value. Wherein the first distance and the second distance may be calculated according to a cosine formula, a euclidean distance formula, or the like. The equipment model is the model of the running equipment corresponding to the equipment training data. The sample types comprise positive samples and negative samples, wherein the positive samples represent that the running equipment corresponding to the equipment training data is in a normal state, and the negative samples represent that the running equipment corresponding to the equipment training data is in a fault state. In one embodiment, the electronic device uses a vector corresponding to both the device model and the sample type as the standard vector. According to the embodiment, the network loss value can be reasonably determined by combining the distance between the equipment training data and the equipment reconstruction data and the distance between the equipment training data and the standard data.
S203, adjusting the first generation network based on the network loss value and a second generation network corresponding to the first generation network to obtain a sample generation model.
In at least one embodiment of the present application, the network structure of the second generation network is similar to or the same as the network structure of the first generation network, the second generation network includes a reconstruction network layer similar to or the same as the network structure of the initial decoder network and a tag identification layer similar to or the same as the network structure of the initial encoder network. The sample generation model is a first generation network that is validated by a second generation network.
In at least one embodiment of the present application, the electronic device adjusts the first generation network based on the network loss value and the second generation network corresponding to the first generation network, and obtaining the sample generation model includes: the electronic equipment adjusts network parameters of the first generation network based on the network loss value to obtain an adjusted first generation network, and inputs equipment training data into the adjusted first generation network to obtain first verification data; the electronic equipment reconstructs the first verification data based on a reconstruction network layer in the second generation network to obtain second verification data, and recognizes the second verification data based on a tag recognition layer in the second generation network to obtain a data tag of the first verification data; the electronic equipment counts the number of the first verification data of which the data label is a preset label as a first number, and counts the total number of the first verification data as a second number. If the ratio of the first quantity to the second quantity is greater than or equal to a preset ratio threshold, the electronic device determines the adjusted first generation network as a sample generation model. If the ratio of the first number to the second number is smaller than the preset ratio threshold, the electronic device readjusts the network parameters of the first generation network.
The network parameters include the number of network layers of the first generation network, and a set value of each network layer, for example, the network parameters include a first weight matrix and a first bias vector in the first coding network layer, and a second weight matrix and a second bias vector in the initial decoder network. The first authentication data is similar to the generation process of the device reconfiguration data. The data tag includes a genuine tag and an imposter tag, and the preset tag is generally set as the genuine tag. The preset ratio threshold may be set for a requirement of the output effect of the sample generation model, where the preset ratio threshold is in positive correlation with the requirement of the output effect of the sample generation model, for example, the higher the requirement of the output effect of the sample generation model, the higher the preset ratio threshold is set.
According to the method and the device, the first generation network can be reasonably adjusted through the network loss value, the fact that the first generation network cannot pass through verification of the second generation network after being adjusted for many times is avoided, adjustment efficiency of the first generation network is improved, verification of the first verification data is achieved through the second generation network, verification of the adjusted first generation network can be achieved, and therefore training effect of the sample generation model is improved.
Specifically, the electronic device adjusts network parameters of the first generation network based on the network loss value, and the obtaining the adjusted first generation network includes: the electronic equipment calculates the ratio of the network loss value to the set value to obtain an adjustment proportion, and adjusts the network parameters based on the adjustment proportion to obtain adjusted parameter values. The electronic device determines an adjusted first generation network based on the network structure of the first generation network and the adjusted parameter value. Wherein, the set value can be set according to the actual situation of the field device, for example, the set value can be set to 100; the set value may also be set to the maximum value of the network parameters. The adjusted parameter value may be obtained after increasing the network parameter based on the adjustment ratio, or may be obtained after decreasing the network parameter based on the adjustment ratio, for example, the adjustment ratio is 0.1, and the network parameter, for example, the network layer number is 100, and the adjusted network layer number may be 110 or 90. According to the embodiment, the adjustment proportion can be reasonably determined by combining the network loss value and the set value, and then the first generation network can be reasonably adjusted based on the adjustment proportion.
Specifically, the electronic device reconstructing the first verification data based on a reconstruction network layer in the second generation network, and obtaining the second verification data includes: the electronic equipment encodes the first verification data based on the encoding layer in the reconstruction network layer to obtain a first verification vector, calculates the first verification vector based on the reconstruction matrix and the reconstruction bias in the reconstruction network layer to obtain an operation vector, and performs reflection processing on the operation vector based on the encoding layer in the reconstruction network layer to obtain second verification data.
Specifically, the electronic device identifying the second verification data based on the tag identification layer in the second generation network, and obtaining the data tag of the first verification data includes: the electronic device encodes the second verification data based on an encoding table in the tag identification layer to obtain a second verification vector, performs normalization processing on the second verification vector to obtain a vector to be analyzed, determines the dimension of which the element value is larger than a preset value from the vector to be analyzed as a first dimension, determines the dimension of which the element value is larger than the preset value in the vector corresponding to the training data of the device based on the tag identification layer as a second dimension, determines that the data tag is a real tag if the first dimension is the same as any second dimension, and determines that the data tag is an impersonation tag if the first dimension is different from a plurality of second dimensions. The preset value is a parameter of the second generation network, the preset value is determined when the second generation network is trained, and the second generation network can be trained according to the learning rate. The vectors corresponding to the device training data are obtained after encoding and normalizing according to the encoding table in the tag identification layer. According to the embodiment, whether the first dimension of the second verification data is the same as the second dimension of the equipment training data is compared, and the accuracy of the data tag is improved because the reference of determining the second dimension is compared with the first dimension.
S204, expanding a fault sample in the acquired data samples based on the sample generation model to obtain expanded data.
In at least one embodiment of the application, the data samples include operating parameter information for an automated device such as a rotating machine, a mechanical electrical device, etc., and the data samples may be the same as or different from the device training data. The fault sample comprises operation parameter information corresponding to the automatic equipment such as rotary mechanical equipment, mechanical and electrical equipment and the like when the automatic equipment fails. And the expansion data is information obtained after the sample generation model reconstructs the fault sample. The generation manner of the extended data is similar to that of the device reconstruction data, and the description thereof will not be repeated.
S205, training a preset detection network based on the data samples and the expansion data to obtain a fault detection model.
In at least one embodiment of the present application, please refer to a network structure schematic diagram of a fault detection model in fig. 4, as shown in fig. 4, the fault detection model includes a feature extraction network and a fault discrimination network, the feature extraction network includes a plurality of extraction networks, and an extraction network corresponding to a device type of a device to be detected is a target extraction network; the fault discrimination network includes a plurality of time convolution layers. The electronic device trains a preset detection network based on the data sample and the expansion data to obtain a fault detection model, and the method comprises the following steps: the electronic equipment inputs the data sample and the expansion data into a preset detection network for detection, an output result is obtained, an identification result of the data sample and the expansion data is obtained, the detection accuracy of the preset detection network is determined according to the output result and the identification result, and the preset detection network is adjusted according to the detection accuracy, so that a fault detection model is obtained. The identification result of the extended data is a fault. The determination of the output result is the same as the determination of the detection result described below, and the present application will not be repeated. According to the embodiment, the preset detection network is trained by combining the expansion data, and the number of the fault samples is expanded to participate in the training of the preset detection network, so that the accuracy of the fault detection model can be improved.
S206, determining parameter characteristics according to the equipment type of the equipment to be tested, and collecting parameter information of the equipment to be tested based on the parameter characteristics.
In at least one embodiment of the present application, the device type includes a device model number and a device class of the device under test. The parameter features are specific parameters corresponding to the type of device, for example, the type of device is a rotating device, and the parameter features are torque, rotational speed, and the like. The parameter information is a parameter value corresponding to the device to be tested in the parameter characteristic.
S207, inputting the equipment type and the parameter information into the fault detection model to obtain a detection result of the equipment to be detected.
In at least one embodiment of the present application, the detection results include a fault result and a normal result. The electronic equipment inputs the equipment type and the parameter information into a fault detection model, and the obtaining of the detection result of the equipment to be detected comprises the following steps: the electronic equipment locates a target extraction network from the feature extraction network based on the equipment type, and performs feature extraction on the parameter information based on the target extraction network to obtain target features. The electronic equipment inputs the target features into a fault discrimination network, obtains discrimination results of a plurality of time convolution layers in the fault discrimination network on the target features, and determines detection results according to the discrimination results of the plurality of time convolution layers on the target features.
The feature extraction network comprises a plurality of extraction networks, different equipment types correspond to different extraction networks, and the target extraction network corresponds to the equipment type of the equipment to be detected. Each temporal convolution layer includes an dilation non-causal convolution and a dilation causal convolution. The most number of detection results are determined, for example, if the determination result a is a fault, the determination result B is a fault, and the determination result C is normal, the detection result is a fault; if the judging result A is normal, the judging result B is fault, the judging result C is normal, and the detecting result is normal.
According to the embodiment, the target extraction network can be positioned through the equipment type, and further the target extraction network is utilized to extract the target characteristics from the parameter information, and the accuracy of the target characteristics can be improved because the target extraction network is adaptive to the equipment type of the equipment to be detected. By combining a plurality of time convolution layers to judge the target characteristics, the target characteristics can be analyzed by utilizing the corresponding time convolution layers under different parameters, so that the detection result can be accurately determined.
According to the technical scheme, the network loss value is determined by combining the first distance and the second distance, and the network loss value can be determined in a plurality of modes, so that the accuracy of the network loss value is improved, and the training effect of the sample generation model is improved; according to the embodiment of the application, the first generation network is verified and adjusted by combining the second generation network, and because the second generation network corresponds to the first generation network, the influence of the network structure on the verification of the first generation network can be eliminated, so that the training effect of the sample generation model can be further improved, the generation rationality and the authenticity of the expansion data are improved, and meanwhile, the second generation network is not required to be constructed when the first generation network is optimized, so that the training efficiency of the sample generation model can be improved; according to the embodiment of the application, the fault samples are expanded through the sample generation model obtained through training, and normal samples in the data samples are not required to be expanded, so that the number of the fault samples can be increased, the problem that the fault detection model cannot be trained to an optimal state due to the fact that the number of the fault samples is too low is avoided, and the detection accuracy of the fault detection model is improved; the embodiment of the application combines the equipment type of the equipment to be detected, can screen out the parameter information related to the detection result for analysis, and can improve the accuracy of the detection result and realize the effective detection of the equipment because the interference of irrelevant data is eliminated and the detection result is determined by using the fault detection model with high detection accuracy.
Fig. 5 is a functional block diagram of an apparatus fault detection device according to an embodiment of the present application. The device fault detection apparatus 11 includes a reconstruction unit 110, a determination unit 111, an adjustment unit 112, an expansion unit 113, a training unit 114, an acquisition unit 115, and an input unit 116. The module/unit referred to herein is a series of computer readable instructions capable of being retrieved by the processor 13 and performing a fixed function and stored in the memory 12.
In an embodiment, the reconstruction unit 110 is configured to reconstruct the acquired device training data based on a first pre-constructed generation network to obtain device reconstruction data; a determining unit 111, configured to determine a network loss value of the first generation network based on a first distance between the device reconstruction data and the device training data, and a second distance between a reconstruction vector corresponding to the device reconstruction data and a standard vector corresponding to the device training data; an adjusting unit 112, configured to adjust the first generation network based on the network loss value and a second generation network corresponding to the first generation network, to obtain a sample generation model; an expansion unit 113, configured to expand a fault sample in the collected data samples based on the sample generation model, so as to obtain expanded data; the training unit 114 is configured to train the preset detection network based on the data samples and the extended data, so as to obtain a fault detection model; the acquisition unit 115 is configured to determine a parameter characteristic according to a device type of the device to be tested, and acquire parameter information of the device to be tested based on the parameter characteristic; and the input unit 116 is used for inputting the equipment type and the parameter information into the fault detection model to obtain a detection result of the equipment to be detected.
In one embodiment, the reconstruction unit 110 is specifically configured to: mapping the equipment training data based on the vector mapping layer to obtain first coding information of the equipment training data; identifying the first coding information based on the first parameter importance identification layer, determining first parameter weight of the equipment training data, and coding the first coding information based on the first coding network layer and the first parameter weight to obtain second coding information of the equipment training data; identifying the second coding information based on the second parameter importance identification layer, determining a second parameter weight of the equipment training data, and coding the second coding information based on the second coding network layer and the second parameter weight to obtain third coding information of the initial coder network to the equipment training data; and mapping the third coding information and the random noise based on the initial decoder network to obtain the device reconstruction data.
In one embodiment, the reconstruction unit 110 is specifically configured to: converting the first coding information into a first input vector of the first coding network layer based on the first parameter weights; according to a first weight matrix and a first bias vector of a first coding network layer, coding a first input vector to obtain characteristic information; and activating the characteristic information according to the activation function of the first coding network layer to obtain second coding information.
In one embodiment, the reconstruction unit 110 is specifically configured to: calculating the element sum of each vector element in the third coding information and the vector element at the corresponding position in the random noise to obtain a second input vector of the initial decoder network; determining an output vector according to a second weight matrix, a second bias vector and a second input vector in the initial decoder network; and determining information corresponding to the output vector as device reconstruction data according to an element mapping table in the initial decoder network.
In one embodiment, the determining unit 111 is specifically configured to: vector encoding is carried out on the equipment reconstruction data to obtain a reconstruction vector, and vector encoding is carried out on the equipment training data to obtain a training vector; calculating the distance between the reconstruction vector and the training vector as a first distance; determining the equipment model corresponding to the equipment training data, and determining a standard vector corresponding to the equipment training data according to the equipment model and the sample type corresponding to the equipment training data; and calculating the distance between the reconstruction vector and the standard vector as a second distance, and carrying out weighted sum operation on the first distance and the second distance to obtain a network loss value.
In one embodiment, the adjusting unit 112 is specifically configured to: based on the network loss value, adjusting network parameters of the first generation network to obtain an adjusted first generation network; inputting the training data of the equipment into the adjusted first generation network to obtain first verification data; reconstructing the first verification data based on a reconstruction network layer in the second generation network to obtain second verification data; identifying the second verification data based on a label identification layer in the second generation network to obtain a data label of the first verification data; the statistical data label is used for counting the number of the first verification data of the preset label as a first number and counting the total number of the first verification data as a second number; if the ratio of the first quantity to the second quantity is greater than or equal to a preset ratio threshold, determining the adjusted first generation network as a sample generation model; or if the ratio of the first number to the second number is smaller than the preset ratio threshold value, readjusting the network parameters of the first generation network.
In one embodiment, the adjusting unit 112 is specifically configured to: calculating the ratio of the network loss value to the set value to obtain an adjustment ratio; adjusting network parameters based on the adjustment proportion to obtain adjusted parameter values; and determining the adjusted first generation network based on the network structure of the first generation network and the adjusted parameter value.
In one embodiment, the input unit 116 is specifically configured to: locating a target extraction network from the feature extraction network based on the device type; extracting the characteristics of the parameter information based on a target extraction network to obtain target characteristics; inputting the target features into a fault discrimination network to obtain discrimination results of a plurality of time convolution layers in the fault discrimination network on the target features; and determining a detection result according to the discrimination results of the plurality of time convolution layers on the target characteristics.
According to the technical scheme, the network loss value is determined by combining the first distance and the second distance, and the network loss value can be determined in a plurality of modes, so that the accuracy of the network loss value is improved, and the training effect of the sample generation model is improved; according to the embodiment of the application, the first generation network is verified and adjusted by combining the second generation network, and because the second generation network corresponds to the first generation network, the influence of the network structure on the verification of the first generation network can be eliminated, so that the training effect of the sample generation model can be further improved, the generation rationality and the authenticity of the expansion data are improved, and meanwhile, the second generation network is not required to be constructed when the first generation network is optimized, so that the training efficiency of the sample generation model can be improved; according to the embodiment of the application, the fault samples are expanded through the sample generation model obtained through training, and normal samples in the data samples are not required to be expanded, so that the number of the fault samples can be increased, the problem that the fault detection model cannot be trained to an optimal state due to the fact that the number of the fault samples is too low is avoided, and the detection accuracy of the fault detection model is improved; the embodiment of the application combines the equipment type of the equipment to be detected, can screen out the parameter information related to the detection result for analysis, and can improve the accuracy of the detection result and realize the effective detection of the equipment because the interference of irrelevant data is eliminated and the detection result is determined by using the fault detection model with high detection accuracy.
The integrated modules/units of the electronic device 1 may be stored in a computer readable storage medium if implemented in the form of software functional units and sold or used as a stand alone product. Based on such understanding, the present application may also be implemented by implementing all or part of the processes in the methods of the embodiments described above, by instructing the associated hardware by means of computer readable instructions, which may be stored in a computer readable storage medium, the computer readable instructions, when executed by a processor, may implement the steps of the respective method embodiments described above.
The computer readable instructions include computer readable instruction code, which may be in the form of source code, object code, executable files, or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer readable instruction code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory).
In particular, the specific implementation method of the processor 13 on the computer readable instructions may refer to the description of the relevant steps in the corresponding embodiment of fig. 2, which is not repeated herein.
In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of modules is merely a logical function division, and other manners of division may be implemented in practice.
The modules illustrated as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional module in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned.
Furthermore, it is evident that the word "comprising" does not exclude other elements or steps, and that the singular does not exclude a plurality. Multiple units or devices may also be implemented by one unit or device in software or hardware. The terms first, second, etc. are used to denote a name, but not any particular order.
Finally, it should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present application and not for limiting the same, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made to the technical solution of the present application without departing from the spirit and scope of the technical solution of the present application.
Claims (10)
1. A method of equipment failure detection, the method comprising:
Reconstructing the acquired device training data based on a first pre-constructed generation network to obtain device reconstruction data;
Determining a network loss value of the first generation network based on a first distance between the equipment reconstruction data and the equipment training data and a second distance between a reconstruction vector corresponding to the equipment reconstruction data and a standard vector corresponding to the equipment training data;
based on the network loss value and a second generation network corresponding to the first generation network, adjusting the first generation network to obtain a sample generation model;
expanding a fault sample in the acquired data sample based on the sample generation model to obtain expanded data;
Training a preset detection network based on the data sample and the expansion data to obtain a fault detection model;
determining parameter characteristics according to the equipment type of equipment to be tested, and acquiring parameter information of the equipment to be tested based on the parameter characteristics;
And inputting the equipment type and the parameter information into the fault detection model to obtain a detection result of the equipment to be detected.
2. The apparatus fault detection method according to claim 1, wherein the first generation network includes an initial encoder network and an initial decoder network, the initial encoder network includes a vector mapping layer, a first parameter importance identification layer, a first encoding network layer, a second parameter importance identification layer, and a second encoding network layer, the reconstructing processing is performed on the acquired apparatus training data based on the pre-constructed first generation network, and obtaining the apparatus reconstruction data includes:
mapping the equipment training data based on the vector mapping layer to obtain first coding information of the equipment training data;
Identifying the first coding information based on the first parameter importance identification layer, determining a first parameter weight of the equipment training data, and coding the first coding information based on the first coding network layer and the first parameter weight to obtain second coding information of the equipment training data;
identifying the second coding information based on the second parameter importance identification layer, determining a second parameter weight of the equipment training data, and coding the second coding information based on the second coding network layer and the second parameter weight to obtain third coding information of the initial encoder network to the equipment training data;
and mapping the third coding information and random noise based on the initial decoder network to obtain the equipment reconstruction data.
3. The apparatus fault detection method according to claim 2, wherein the encoding the first encoded information based on the first encoded network layer and the first parameter weight, to obtain the second encoded information of the apparatus training data includes:
converting the first coding information into a first input vector of the first coding network layer based on the first parameter weights;
coding the first input vector according to a first weight matrix and a first bias vector of the first coding network layer to obtain characteristic information;
and activating the characteristic information according to the activation function of the first coding network layer to obtain the second coding information.
4. The apparatus fault detection method according to claim 2, wherein the mapping the third encoded information and random noise based on the initial decoder network to obtain the apparatus reconstruction data comprises:
Calculating the element sum of each vector element in the third coding information and the vector element at the corresponding position in the random noise to obtain a second input vector of the initial decoder network;
determining an output vector according to a second weight matrix, a second bias vector and the second input vector in the initial decoder network;
and determining information corresponding to the output vector as the device reconstruction data according to an element mapping table in the initial decoder network.
5. The apparatus fault detection method according to claim 1, wherein the determining the network loss value of the first generation network based on a first distance between the apparatus reconstruction data and the apparatus training data, and a second distance between a reconstruction vector corresponding to the apparatus reconstruction data and a standard vector corresponding to the apparatus training data comprises:
Vector encoding is carried out on the equipment reconstruction data to obtain the reconstruction vector, and vector encoding is carried out on the equipment training data to obtain a training vector;
calculating the distance between the reconstruction vector and the training vector as the first distance;
determining a device model corresponding to the device training data, and determining a standard vector corresponding to the device training data according to the device model and a sample type corresponding to the device training data;
And calculating the distance between the reconstruction vector and the standard vector as the second distance, and carrying out weighted sum operation on the first distance and the second distance to obtain the network loss value.
6. The apparatus fault detection method according to claim 1, wherein the adjusting the first generation network based on the network loss value and a second generation network corresponding to the first generation network, to obtain a sample generation model includes:
Based on the network loss value, adjusting network parameters of the first generation network to obtain an adjusted first generation network;
Inputting the training data of the equipment into the adjusted first generation network to obtain first verification data;
reconstructing the first verification data based on a reconstruction network layer in the second generation network to obtain second verification data;
Identifying the second verification data based on a label identification layer in the second generation network to obtain a data label of the first verification data;
Counting the number of first verification data of which the data tag is a preset tag as a first number, and counting the total number of the first verification data as a second number;
if the ratio of the first quantity to the second quantity is greater than or equal to a preset ratio threshold, determining the adjusted first generation network as the sample generation model; or alternatively
And if the ratio of the first quantity to the second quantity is smaller than the preset ratio threshold value, readjusting the network parameters of the first generation network.
7. The method of claim 6, wherein adjusting network parameters of the first generation network based on the network loss value, the adjusted first generation network comprising:
calculating the ratio of the network loss value to the set value to obtain an adjustment ratio;
Adjusting the network parameters based on the adjustment proportion to obtain adjusted parameter values;
And determining the adjusted first generation network based on the network structure of the first generation network and the adjusted parameter value.
8. The method for detecting a device fault according to claim 1, wherein the fault detection model includes a feature extraction network and a fault discrimination network, and the inputting the device type and the parameter information into the fault detection model, obtaining a detection result of the device to be detected includes:
locating a target extraction network from the feature extraction networks based on the device type;
performing feature extraction on the parameter information based on the target extraction network to obtain target features;
inputting the target features into the fault discrimination network to obtain discrimination results of a plurality of time convolution layers in the fault discrimination network on the target features;
and determining the detection result according to the discrimination results of the plurality of time convolution layers on the target features.
9. An electronic device, comprising:
A memory storing computer readable instructions; and
A processor executing computer readable instructions stored in the memory to implement the device fault detection method of any one of claims 1 to 8.
10. A computer readable storage medium having stored therein computer readable instructions that are executed by a processor in an electronic device to implement the device failure detection method of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410427341.0A CN118035730B (en) | 2024-04-10 | Equipment fault detection method, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410427341.0A CN118035730B (en) | 2024-04-10 | Equipment fault detection method, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118035730A true CN118035730A (en) | 2024-05-14 |
| CN118035730B CN118035730B (en) | 2024-06-28 |
Family
ID=
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021174723A1 (en) * | 2020-03-02 | 2021-09-10 | 平安科技（深圳）有限公司 | Training sample expansion method and apparatus, electronic device, and storage medium |
| CN113627576A (en) * | 2021-10-08 | 2021-11-09 | 平安科技（深圳）有限公司 | Code scanning information detection method, device, equipment and storage medium |
| CN116206171A (en) * | 2022-12-08 | 2023-06-02 | 国能铁路装备有限责任公司 | Method and device for detecting foreign matter faults of vehicle, electronic equipment and storage medium |
| CN116452802A (en) * | 2023-03-22 | 2023-07-18 | 平安科技（深圳）有限公司 | Vehicle loss detection method, device, equipment and storage medium |
| WO2023231374A1 (en) * | 2022-05-30 | 2023-12-07 | 深圳技术大学 | Semi-supervised fault detection and analysis method and apparatus for mechanical device, terminal, and medium |
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021174723A1 (en) * | 2020-03-02 | 2021-09-10 | 平安科技（深圳）有限公司 | Training sample expansion method and apparatus, electronic device, and storage medium |
| CN113627576A (en) * | 2021-10-08 | 2021-11-09 | 平安科技（深圳）有限公司 | Code scanning information detection method, device, equipment and storage medium |
| WO2023231374A1 (en) * | 2022-05-30 | 2023-12-07 | 深圳技术大学 | Semi-supervised fault detection and analysis method and apparatus for mechanical device, terminal, and medium |
| CN116206171A (en) * | 2022-12-08 | 2023-06-02 | 国能铁路装备有限责任公司 | Method and device for detecting foreign matter faults of vehicle, electronic equipment and storage medium |
| CN116452802A (en) * | 2023-03-22 | 2023-07-18 | 平安科技（深圳）有限公司 | Vehicle loss detection method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jack et al. | Genetic algorithms for feature selection in machine condition monitoring with vibration signals | |
| CN113673346B (en) | Motor vibration data processing and state identification method based on multiscale SE-Resnet | |
| KR101964412B1 (en) | Method for diagnosing anomaly log of mobile commmunication data processing system and system thereof | |
| Gangsar et al. | Diagnostics of mechanical and electrical faults in induction motors using wavelet-based features of vibration and current through support vector machine algorithms for various operating conditions | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| Kumar et al. | The Importance of Feature Processing in Deep‐Learning‐Based Condition Monitoring of Motors | |
| CN116168720A (en) | Motor sound abnormality fault diagnosis method, system and storable medium | |
| CN116578843A (en) | Centrifugal pump diagnostic model training method, diagnostic method, system, device and medium | |
| CN114048787A (en) | Method and system for intelligently diagnosing bearing fault in real time based on Attention CNN model | |
| CN117036732B (en) | Electromechanical equipment detection system, method and equipment based on fusion model | |
| Senanayaka et al. | A robust method for detection and classification of permanent magnet synchronous motor faults: Deep autoencoders and data fusion approach | |
| CN118035730B (en) | Equipment fault detection method, electronic equipment and storage medium | |
| CN116150186B (en) | Communication Internet of things management method and system based on big data | |
| CN116907029A (en) | Method for detecting abnormality of fan in outdoor unit, control device and air conditioner outdoor unit | |
| CN115356599B (en) | Multi-mode urban power grid fault diagnosis method and system | |
| CN117113078A (en) | Small sample bearing fault mode identification method and system based on multi-source data integration | |
| CN116541771A (en) | Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion | |
| CN118035730A (en) | Equipment fault detection method, electronic equipment and storage medium | |
| CN116595857A (en) | Rolling bearing multistage degradation residual life prediction method based on deep migration learning | |
| CN115982621A (en) | Rotary machine residual service life prediction method based on time convolution network | |
| CN115248377A (en) | Asynchronous motor rotor broken bar fault detection method, computer device and medium | |
| US20220187789A1 (en) | Equipment failure diagnosis apparatus, equipment failure diagnosis method, smart factory system and application agent | |
| Hao et al. | New fusion features convolutional neural network with high generalization ability on rolling bearing fault diagnosis | |
| CN114997333A (en) | Fault diagnosis method and device for wind driven generator | |
| CN113505817A (en) | Self-adaptive weighting training method for bearing fault diagnosis model samples under unbalanced data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |