Equipment fault prediction method and device based on industrial internet, and storage medium

Abstract

The invention discloses an industrial internet-based equipment fault prediction method and device and a storage medium, and relates to the technical field of industrial internet, and the method comprises the steps: collecting the target operation data of equipment in real time; the target operation data is preprocessed and abnormal signal identification is carried out to obtain an identification result, so that potential fault hidden dangers can be found in time, major accidents caused by small problems which are not found in time are effectively prevented, and the overall reliability and stability of equipment are enhanced; when the identification result is that the abnormal signal is detected, the fault type and the transfer path of the abnormal signal are predicted based on the hybrid model and the knowledge graph to obtain a prediction result, so that the fault type corresponding to the abnormal signal and the possible development path thereof can be predicted more accurately; and generating a control instruction based on the prediction result and feeding back to the equipment end. Compared with a traditional single-model diagnosis mode, more accurate and comprehensive fault information can be provided, a targeted maintenance strategy can be rapidly formulated, and downtime is shortened.

Description

Technical Field

[0001] The present application relates to the technical field of industrial Internet, and particularly to a method, device and storage medium for equipment fault prediction based on the industrial Internet. Background Art

[0002] With the continuous improvement of the level of industrial automation and informatization, the complexity and refinement degree of industrial equipment are also getting higher and higher, and the association between each component is closer. In this case, even potential hidden dangers or minor faults may lead to the failure of the entire equipment, and even cause greater catastrophic damage, with extremely serious consequences. Therefore, the role of equipment fault prediction technology is becoming more and more important. It can detect existing or upcoming faults in equipment early, quickly locate the cause of the fault, give a treatment plan and predict the development trend of the fault, or prevent upcoming equipment safety threats, thereby greatly reducing the time for fault discovery and repair, improving the repair quality, saving the repair cost, and further improving the availability and safety of the equipment.

[0003] Currently, the equipment fault prediction can draw on traditional communication network management methods, but there is still a problem of low fault prediction accuracy. Summary of the Invention

[0004] The main purpose of the present application is to provide a method, device and storage medium for equipment fault prediction based on the industrial Internet, aiming to solve the technical problem of low fault prediction accuracy.

[0005] To achieve the above object, the present application proposes a method for equipment fault prediction based on the industrial Internet, and the method for equipment fault prediction based on the industrial Internet includes:

[0006] Real-time collect the target operation data of the equipment;

[0007] Preprocess the target operation data and identify abnormal signals to obtain an identification result;

[0008] When the identification result is that an abnormal signal is detected, predict the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result;

[0009] Generate a control instruction based on the prediction result and feedback it to the equipment side.

[0010] In an embodiment, the step of, when the identification result is that an abnormal signal is detected, predicting the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result includes:

[0011] When the identification result is that an abnormal signal is detected, input the operation data with anomalies into the hybrid model for fault prediction to obtain a fault probability;

[0012] Determine the fault type of the abnormal signal according to the fault probability;

[0013] Based on the knowledge graph to associate historical fault cases, predict the transfer path of the abnormal signal to obtain the predicted fault transfer path;

[0014] Obtain the prediction result through the fault type and the predicted fault transfer path.

[0015] In one embodiment, the step of inputting the abnormal operation data into the hybrid model for fault prediction to obtain the fault probability when the recognition result is that an abnormal signal is detected includes:

[0016] When the recognition result is that an abnormal signal is detected, divide the abnormal operation data according to a preset time window to obtain a plurality of subsequences, and the subsequences are expressed as: X = {x1, x2,..., xT}, where X represents the data sequence within a time window, T is the time step, and xt is the operation data at the t-th moment;

[0017] Input the plurality of subsequences into the convolutional neural network in the hybrid model for local feature extraction to obtain local features;

[0018] Perform feature splicing on the local features through the bidirectional LSTM in the hybrid model to obtain spliced features;

[0019] Perform fault prediction on the spliced features through the fully connected layer in the hybrid model to obtain the fault probability within a preset time period.

[0020] In one embodiment, the step of preprocessing the target operation data and identifying the abnormal signal to obtain the recognition result includes:

[0021] Perform data cleaning, normalization, and feature extraction on the target operation data to obtain the preprocessed target operation data;

[0022] Use the generative adversarial network model to generate an updated data set for the preprocessed target operation data;

[0023] Add the updated data set to the preprocessed target operation data to form a balanced data set;

[0024] Perform abnormal signal recognition on the balanced data set to obtain the recognition result.

[0025] In one embodiment, the step of performing abnormal signal recognition on the balanced data set to obtain the recognition result includes:

[0026] Reconstruct the balanced data set through a preset autoencoder model to obtain reconstructed output data;

[0027] Calculate the reconstruction error between the balanced data set and the reconstructed output data;

[0028] Compare the reconstruction error with a preset error threshold to identify abnormal signals;

[0029] When the reconstruction error is greater than the preset error threshold, determine that the recognition result is that an abnormal signal is detected.

[0030] In one embodiment, the step of acquiring the target operation data of the real-time acquisition device includes:

[0031] Real-time obtain multi-dimensional operation data through a deployed sensor network, and the multi-dimensional operation data at least includes temperature, voltage, vibration and noise data;

[0032] Perform preliminary anomaly detection on the multi-dimensional operation data at the edge computing node to obtain a preliminary detection result;

[0033] Use the multi-dimensional operation data without anomalies in the preliminary detection result as the target operation data.

[0034] In one embodiment, the step of generating a control instruction based on the prediction result and feeding it back to the device side includes:

[0035] Determine the fault probability, predicted fault transfer path and fault type based on the prediction result;

[0036] When the fault probability exceeds a preset probability threshold, generate a control instruction according to the fault type and the predicted fault transfer path, and feed the control instruction back to the device side.

[0037] In one embodiment, the step of generating a control instruction based on the prediction result and feeding it back to the device side includes:

[0038] Obtain the vibration amplitude and noise change rate in the target operation data according to the prediction result;

[0039] When the vibration amplitude is the same as the seismic wave data and the noise change rate is within a preset range, detect the Internet seismic data;

[0040] When the Internet seismic data indicates an earthquake, generate a shutdown control instruction;

[0041] Feed the shutdown control instruction back to the device side so that the device side performs a shutdown process after a preset time based on the shutdown control instruction.

[0042] In addition, to achieve the above object, the present application also proposes a device fault prediction device based on the industrial Internet, and the device fault prediction device based on the industrial Internet includes:

[0043] An acquisition module, configured to collect target operation data of the device in real time;

[0044] A processing and recognition module, configured to perform preprocessing and abnormal signal recognition on the target operation data to obtain a recognition result;

[0045] A prediction module, configured to, when the recognition result is that an abnormal signal is detected, predict the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result;

[0046] A generation module, configured to generate a control instruction based on the prediction result and feedback it to the device side.

[0047] In addition, to achieve the above object, the present application also proposes a device fault prediction device based on the industrial Internet. The device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. The computer program is configured to implement the steps of the device fault prediction method based on the industrial Internet as described above.

[0048] In addition, to achieve the above object, the present application also proposes a storage medium. The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the device fault prediction method based on the industrial Internet as described above.

[0049] In addition, to achieve the above object, the present application also provides a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, it implements the steps of the device fault prediction method based on the industrial Internet as described above.

[0050] One or more technical solutions proposed by the present application collect target operation data of the device in real time; perform preprocessing and abnormal signal recognition on the target operation data to obtain a recognition result, which can timely discover potential fault hazards and effectively prevent major accidents caused by small problems that are not discovered in time, thereby enhancing the overall reliability and stability of the device; when the recognition result is that an abnormal signal is detected, predict the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result, which can more accurately predict the fault type corresponding to the abnormal signal and its possible development path; generate a control instruction based on the prediction result and feedback it to the device side. Compared with the traditional single-model diagnosis method, it can provide more accurate and comprehensive fault information, which helps to quickly formulate targeted maintenance strategies and reduce downtime. Brief Description of the Drawings

[0051] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with this application, and are used together with the specification to explain the principles of this application.

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings required for use in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts.

[0053] Figure 1 It is a schematic flowchart provided for the first embodiment of the device fault prediction method based on the industrial Internet in this application;

[0054] Figure 2 It is a schematic flowchart provided for the second embodiment of the device fault prediction method based on the industrial Internet in this application;

[0055] Figure 3 It is a schematic flowchart provided for the third embodiment of the device fault prediction method based on the industrial Internet in this application;

[0056] Figure 4 It is a schematic flowchart provided for the fourth embodiment of the device fault prediction method based on the industrial Internet in this application;

[0057] Figure 5 It is a schematic module structure diagram of the device fault prediction device based on the industrial Internet for the embodiments of this application;

[0058] Figure 6 It is a schematic device structure diagram of the hardware operating environment involved in the device fault prediction method based on the industrial Internet for the embodiments of this application.

[0059] The realization of the purpose, functional features and advantages of this application will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0060] It should be understood that the specific embodiments described here are only used to explain the technical solutions of this application and are not used to limit this application.

[0061] To better understand the technical solutions of this application, the following will be described in detail in combination with the specification drawings and specific embodiments.

[0062] The main solution of the embodiment of this application is: real-time collect multi-dimensional operation data of the device, such as temperature, voltage, vibration, noise, etc., preprocess the collected data and identify abnormal signals, determine whether abnormal signals are detected, if so, predict the fault type and transfer path based on the hybrid model (CNN + LSTM) and knowledge graph, and generate control instructions to feedback to the device side.

[0063] Since the existing technology mainly collects in advance the device data that generates alarms, thereby training an alarm model and directly outputting the fault detection result through the alarm model, but it cannot predict the device fault in advance, resulting in inaccurate detection results.

[0064] This application provides a solution, which includes: real-time collect the target operation data of the device; preprocess the target operation data and identify abnormal signals to obtain an identification result; when the identification result is that abnormal signals are detected, predict the fault type and transfer path of the abnormal signals based on the hybrid model and knowledge graph to obtain a prediction result; generate control instructions based on the prediction result and feedback them to the device side.

[0065] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication, and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device that can implement the above functions, a device fault prediction device based on the industrial Internet, etc. Hereinafter, a device fault prediction device based on the industrial Internet will be taken as an example to illustrate this embodiment and the following embodiments.

[0066] Based on this, the embodiment of this application provides a device fault prediction method based on the industrial Internet, referring to Figure 1 , Figure 1 which is a schematic flowchart of the first embodiment of the device fault prediction method based on the industrial Internet of this application.

[0067] In this embodiment, the device fault prediction method based on the industrial Internet includes steps S10 to S40:

[0068] Step S10: Real-time collect the target operation data of the device.

[0069] It should be noted that the target operation data of the device is the operation data of the collected device after screening, such as temperature data, voltage data, vibration data, noise data, etc., and may also include other data, which is not limited in this embodiment.

[0070] In specific implementation, the operation data of the device can be real-time collected by deploying a sensor network, and the collected data can be preliminarily processed to obtain the target operation data.

[0071] In a feasible implementation, step S10 may include steps A11 to A13:

[0072] Step A11: Obtain multi-dimensional operation data in real time through a deployed sensor network, where the multi-dimensional operation data includes at least temperature, voltage, vibration, and noise data.

[0073] In a specific implementation, a deployed sensor network can be used to collect multi-dimensional operation data of a device in real time. The sensors can include temperature sensors, voltage sensors, vibration sensors, and noise sensors, etc. The number of sensors can be adjusted according to the number of devices. The temperature data, voltage data, vibration data, and noise data of the device are collected in real time through each sensor.

[0074] It should be noted that after obtaining the multi-dimensional data, an Internet of Things gateway supporting industrial protocols can be used for real-time data transmission. The industrial protocols can be Modbus, OPC UA, etc.

[0075] Step A12: Perform preliminary anomaly detection on the multi-dimensional operation data at an edge computing node to obtain a preliminary detection result.

[0076] When the device fault prediction device receives the data transmitted by the Internet of Things gateway, preliminary anomaly detection can be performed on the multi-dimensional operation data at the edge computing node to obtain a preliminary detection result. Since each device is connected to the nearest edge computing node, these nodes perform preliminary data analysis and anomaly detection, thereby reducing the data transmission volume and improving the response speed.

[0077] For example, the multi-dimensional operation data is represented as D(t) = {d1, d2,..., dn}, where D(t) represents a set of sensor data at time t. By preprocessing D(t), including filtering and normalization, and then extracting a feature vector that can represent the data characteristics, the feature vector F is represented as follows:

[0078] F = {f1, f2,..., fm} = g(D(t))

[0079] where g(.) represents a function for extracting features from the data, and F is the finally obtained feature vector.

[0080] In a specific implementation, a trained scoring model can be used to calculate the anomaly score of the feature vector, and this score reflects the degree to which the data deviates from the normal mode.

[0081] S = h(F, M)

[0082] where S is the anomaly score and h(.) is a function for calculating the anomaly score.

[0083] Through the above formula calculation, the anomaly scores of each data are obtained, and the scores are compared with the score threshold T. If the anomaly score exceeds the score threshold T, the data is marked as abnormal; otherwise, it is regarded as normal. By marking all multi-dimensional operation data, a preliminary detection result with marks is obtained.

[0084] Step A13: Use the multi-dimensional operation data without anomalies in the preliminary detection result as the target operation data.

[0085] In a specific implementation, the data marked as abnormal in the preliminary detection result can be excluded from the multi-dimensional operation data, and the multi-dimensional operation data marked as normal can be used as the target operation data.

[0086] Step S20: Perform preprocessing and abnormal signal recognition on the target operation data to obtain a recognition result.

[0087] It should be noted that after obtaining the target operation data, the target operation data can be preprocessed again, such as data cleaning, normalization, and feature extraction, etc., to obtain preprocessed data, and abnormal signal recognition is performed on the preprocessed data, so as to judge whether there is signal abnormality in the corresponding data, quickly judge whether there is a potential fault in the device, and obtain a recognition result.

[0088] The recognition result includes detecting an abnormal signal or not detecting an abnormal signal.

[0089] If the recognition result is not detecting an abnormal signal, the operation data of the device can be continuously collected and processed, and abnormal signal recognition is performed again.

[0090] Step S30: When the recognition result is detecting an abnormal signal, predict the fault type and transfer path of the abnormal signal based on the hybrid model and the knowledge graph to obtain a prediction result.

[0091] It should be noted that if the recognition result is detecting an abnormal signal, it is determined that the device corresponding to this data may have a fault. Therefore, the fault type of the device detecting the abnormal signal can be predicted based on the hybrid model and the knowledge graph to obtain a specific prediction result.

[0092] The hybrid model includes a CNN and a bidirectional LSTM model. The local features of the device's corresponding operation data are extracted by the CNN, and the temporal dependencies of the data are captured by the bidirectional LSTM model, and then the fusion features of the data are obtained, and fault prediction is performed based on the fusion features to determine the fault type of the device.

[0093] The knowledge graph can be associated with historical fault cases, so that the fault transfer path of the device can be predicted based on the historical fault cases. For example, if the device is a power supply device and the fault type of the device is local overheating, the fault transfer path is from local overheating to circuit board damage.

[0094] Step S40: Generate a control instruction based on the prediction result and feedback it to the device side.

[0095] In a specific implementation, corresponding control instructions can be generated according to the specific prediction result, and then the corresponding control instructions are fed back to the device side, so that the device side performs corresponding operations based on this control instruction. For example, if a device shutdown instruction is generated according to the prediction result, the device shutdown instruction can be fed back to the device side to control the device to shut down, or if a device restart instruction is generated according to the prediction result, the device restart instruction is fed back to the device side to control the device to restart. Or a control instruction for reducing the output voltage and notifying the maintenance personnel to check the cooling system is generated according to the prediction result, so as to process the device through this control instruction.

[0096] In a specific implementation, after the device performs the corresponding operation, the running state of the device can also be monitored in real time, and then it is determined whether the fault is solved. If the fault is not solved, return to step S30 to optimize the model parameters of the hybrid model and perform subsequent predictions.

[0097] This embodiment provides a device fault prediction method based on the industrial Internet, which collects the target running data of the device in real time; preprocesses the target running data and identifies abnormal signals to obtain an identification result, which can timely discover potential fault hazards and effectively prevent major accidents caused by small problems that are not discovered in time, thereby enhancing the overall reliability and stability of the device; when the identification result is that an abnormal signal is detected, based on the hybrid model and the knowledge graph, predict the fault type and transfer path of the abnormal signal to obtain a prediction result, which can more accurately predict the fault type corresponding to the abnormal signal and its possible development path; generate a control instruction based on the prediction result and feedback it to the device side. Compared with the traditional single-model diagnosis method, it can provide more accurate and comprehensive fault information, which helps to quickly formulate targeted maintenance strategies and reduce downtime.

[0098] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar content as in the above-mentioned first embodiment can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 2 , step S30 includes steps S301 to S304:

[0099] Step S301: When the identification result is that an abnormal signal is detected, input the running data with abnormalities into the hybrid model for fault prediction to obtain a fault probability.

[0100] It should be noted that if the recognition result is the detection of an abnormal signal, the operation data corresponding to the device with the abnormal signal can be input into the hybrid model for fault prediction, so as to obtain the fault probability of the predicted fault. The fault probability can be 30%, 80%, 90%, and the fault probability reflects the possibility of the device having a fault.

[0101] In a feasible implementation manner, step S401 may include steps B11 to B14:

[0102] Step B11: When the recognition result is the detection of an abnormal signal, the abnormal operation data is segmented according to a preset time window to obtain a plurality of subsequences.

[0103] The subsequence is expressed as: X = {x1, x2,..., xT}, where X represents the data sequence within a time window, T is the time step, and xt is the operation data at the t-th moment.

[0104] It should be noted that if the recognition result is the detection of an abnormal signal, it is possible to further determine whether there is a fault in the corresponding device. Therefore, the abnormal operation data can be segmented according to a preset time window. The preset time window can be set to 5 minutes, 8 minutes, etc., and this embodiment does not limit this.

[0105] It should be noted that, for example, if the preset time window is 5 minutes, the abnormal operation data is segmented according to a time window of every 5 minutes, so as to obtain a plurality of subsequences. The time step of T is 5, so as to obtain a plurality of subsequences.

[0106] Step B12: Input the plurality of subsequences into the convolutional neural network in the hybrid model for local feature extraction to obtain local features.

[0107] In a specific implementation, the hybrid model is provided with a convolutional neural network CDD and a bidirectional LSTM. The convolutional neural network is a one-dimensional convolutional neural network. By using the one-dimensional convolutional neural network to extract local features from the data in each time window, local features are obtained.

[0108] It can be understood that assuming 3 convolutional filters are used, and the size of each filter is k, the convolution process is as follows:

[0109] Fconv = Conv(X, W)

[0110] Where W is the convolution kernel weight matrix, Conv(.) represents the convolution operation, and Fconv is the feature map after convolution, that is, the local feature.

[0111] Step B13: Use the bidirectional LSTM in the hybrid model to splice the local features to obtain spliced features.

[0112] It should be noted that in order to capture long-term dependencies such as the temperature rise trend, bidirectional LSTM can be applied to further process the feature sequence output by CNN, so as to splice local features and obtain the final spliced features.

[0113] The forward and backward LSTMs process local features respectively:

[0114] F forward = LSTM forward (Fconv)

[0115] F backward = LSTM backward (Fconv)

[0116] And splice the forward and backward information, F context = Concatenate([F forward , F backward ), where Concatenate(.) means splicing the features in two directions to obtain the spliced features.

[0117] Step B14: Perform fault prediction on the spliced features through the fully connected layer in the hybrid model to obtain the fault probability within a preset time period.

[0118] In specific implementation, one or more fully connected layers can be used to map the spliced features to the fault probability space, so as to predict the fault probability within a future time period.

[0119] P fault = Softmax(Dense(F context ))

[0120] where Softmax(.) is used to convert the output into a probability distribution, and Dense(.) represents the fully connected layer operation. P fault is the predicted fault probability distribution.

[0121] Step S302: Determine the fault type of the abnormal signal according to the fault probability.

[0122] It should be noted that the fault type of the abnormal signal can be determined according to the determined fault probability. According to the calculated probability distribution, the fault type with the highest probability is selected as the fault type of the current abnormal signal. If there are multiple fault types with similar probability values, professional knowledge can be further combined or additional diagnostic tests can be taken for judgment.

[0123] Step S303: Based on the knowledge graph to associate historical fault cases, predict the transfer path of the abnormal signal to obtain the predicted fault transfer path.

[0124] It should be noted that historical fault cases can be associated based on the knowledge graph. Specifically, historical fault cases can be determined according to the type of equipment. For example, if the equipment is a manipulator, the historical fault cases are historical bearing wear, motor overheating and other fault cases. If the equipment is a power supply device, the historical fault cases are historical overload cases, historical overheating cases, etc.

[0125] In this embodiment, a knowledge graph including the relationships between equipment components, fault types and their transition probabilities can be constructed according to historical fault cases and the experience of domain experts, so as to predict the transfer path of abnormal signals based on the knowledge graph and obtain the predicted fault transfer path. For example, the predicted fault transfer path is the path from vibration abnormality to displacement deviation.

[0126] Step S304: Obtain a prediction result through the fault type and the predicted fault transfer path.

[0127] It can be understood that the predicted fault type of the equipment and the predicted fault transfer path can be used as the final fault prediction result of the equipment.

[0128] In this embodiment, when the recognition result is that an abnormal signal is detected, the abnormal operation data is input into the hybrid model for fault prediction to obtain a fault probability; the fault type of the abnormal signal is determined according to the fault probability; historical fault cases are associated based on the knowledge graph, and the transfer path of the abnormal signal is predicted to obtain a predicted fault transfer path; a prediction result is obtained through the fault type and the predicted fault transfer path. By foreseeing the potential fault development path in advance, it helps to formulate a more scientific and reasonable preventive maintenance plan and reduce the unexpected downtime.

[0129] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar content as in the above-mentioned first embodiment can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 3 , step S20 includes steps S201 to S204:

[0130] Step S201: Perform data cleaning, normalization and feature extraction on the target operation data to obtain the preprocessed target operation data.

[0131] It should be noted that after obtaining the target operation data of the equipment, the target operation data can be preprocessed first, specifically including data cleaning, normalization processing and feature extraction processing, so as to obtain the preprocessed target operation data. For example, data such as the operation coefficient of the vibration amplitude, the temperature change rate and the noise change rate of the equipment can be calculated.

[0132] The operation coefficient of the vibration amplitude is calculated as follows:

[0133] Rv = Amax / Arms

[0134] Where Amax is the maximum vibration amplitude and Arms is the root mean square vibration amplitude.

[0135] Step S202: Generate an updated data set from the preprocessed target operation data using a generative adversarial network model.

[0136] In a specific implementation, the generative adversarial network model can be a WGAN-GP model or other types of models. The WGAN-GP model can be obtained through preliminary training. Different types of samples can be collected as the training set, and then the model can be trained through multiple rounds of iteration to obtain the generative adversarial network model. The model can be evaluated using a predetermined quality assessment criterion, such as indicators like similarity to real samples and classification accuracy, and the model parameters can be adjusted according to the evaluation results to obtain the final generative adversarial network model.

[0137] In a specific implementation, the preprocessed target operation data can be input into the generative adversarial network model to generate an updated data set, which is new generated data. The generative model can create more representative synthetic samples based on limited real samples, effectively alleviating the problems caused by insufficient samples, thereby expanding the preprocessed target operation data.

[0138] Step S203: Add the updated data set to the preprocessed target operation data to form a balanced data set.

[0139] In a specific implementation, the updated data set can be added to the preprocessed target operation data to form a balanced data set.

[0140] Step S204: Identify abnormal signals in the balanced data set to obtain an identification result.

[0141] In a specific implementation, abnormal signals in the balanced data set can be identified to obtain an identification result on whether there are abnormal signals.

[0142] In a feasible implementation manner, step S204 can include steps C11 to C14:

[0143] Step C11: Reconstruct the balanced data set through a preset autoencoder model to obtain reconstructed output data.

[0144] It should be noted that the preset autoencoder model can be an autoencoder model obtained through prior training. By inputting the balanced data set into the preset autoencoder model for reconstruction, reconstructed output data can be obtained.

[0145] Step C12: Calculate the reconstruction error between the balanced data set and the reconstructed output data.

[0146] In a specific implementation, an error metric can be performed on the error between the balanced dataset and the reconstructed output data. The error metric methods can include mean square error, mean absolute error, etc. For example, the reconstruction error is calculated by mean square error as follows:

[0147]

[0148] where, x i represents the i-th element in the balanced dataset, represents the corresponding element of the output, and n is the total number of elements.

[0149] Step C13: Compare the reconstruction error with a preset error threshold to identify abnormal signals.

[0150] In a specific implementation, the preset error threshold can be set in advance. For example, a suitable threshold is selected by analyzing the distribution of reconstruction errors in the normal dataset. For each input signal, after calculating its reconstruction error, it is compared with the preset error threshold.

[0151] Step C14: When the reconstruction error is greater than the preset error threshold, determine that the recognition result is an abnormal signal detected.

[0152] If the reconstruction error is greater than the preset error threshold, then the signal is considered an abnormal signal; otherwise, it is regarded as normal.

[0153] In this embodiment, data cleaning, normalization, and feature extraction are performed on the target operation data to obtain the preprocessed target operation data; the preprocessed target operation data is used to generate an updated dataset by using a generative adversarial network model; the updated dataset is added to the preprocessed target operation data to form a balanced dataset; abnormal signal recognition is performed on the balanced dataset to obtain a recognition result. The generative adversarial network model is used to generate an updated dataset according to the preprocessed target operation data. This not only increases the amount of data, but also, due to the unique properties of the generative adversarial network model, can synthesize representative samples, especially for minority class samples, which helps to solve the class imbalance problem. The balanced dataset improves the robustness and generalization ability of the abnormal signal recognition algorithm and reduces the bias caused by uneven data distribution; based on the balanced dataset for abnormal signal recognition, the subtle differences between normal and abnormal states can be captured more accurately, significantly improving the sensitivity and specificity of abnormal detection.

[0154] Based on the first embodiment of the present application, in the fourth embodiment of the present application, the same or similar content as in the above-mentioned embodiment 1 can be referred to the above introduction and will not be elaborated hereinafter. On this basis, please refer to Figure 4, step S40 includes steps S401 to S402:

[0155] Step S401: Determine the failure probability, predicted failure transfer path, and failure type based on the prediction result.

[0156] It should be noted that after obtaining the prediction result, the failure probability, failure type, and predicted failure transfer path of the device can be determined according to the specific prediction result.

[0157] Step S402: When the failure probability exceeds the preset probability threshold, generate a control instruction according to the failure type and the predicted failure transfer path, and feedback the control instruction to the device side.

[0158] In a specific implementation, a preset probability threshold can be set. If the failure probability exceeds this preset probability threshold, it indicates that the device is about to fail. If the failure probability does not exceed this preset probability threshold, the device can continue to be predicted for failures. For example, the preset probability threshold is set to 90%. When the failure probability exceeds 90%, corresponding control instructions can be generated according to the failure type and the predicted failure transfer path, and the control instructions are feedback to the device side to control the device side.

[0159] In a feasible implementation manner, step S40 may further include: obtaining the vibration amplitude and noise change rate in the target operation data according to the prediction result; detecting the Internet earthquake data when the vibration amplitude is the same as the seismic wave data and the noise change rate is within a preset range; generating a shutdown control instruction when the Internet earthquake data indicates an earthquake; and feedbacking the shutdown control instruction to the device side so that the device side performs a shutdown process based on the shutdown control instruction after a preset duration.

[0160] It can monitor the vibration amplitude of the device to monitor whether there is an earthquake, and perform corresponding processing on the device in time before the earthquake to avoid device damage. Therefore, the vibration amplitude and noise change rate of the device can be obtained. If the vibration amplitude is similar to the seismic wave data and the noise change rate of the device is within the preset range, that is, the stable value, the Internet seismic data can be queried in real time. If there is an earthquake in the Internet seismic data, a shutdown control instruction can be generated, and it can be prompted on the display terminal of the device that the device will stop after 1 minute, giving early warnings to all operators. After 1 minute, all devices are shut down, not directly stopped, which can not only ensure the safety of personnel and equipment, but also provide the on-site management personnel with time to cancel the command, avoid outputting wrong commands due to wrong information, and prevent damage to the device caused by problems such as circuit short-circuit and leakage due to the earthquake. If the noise change rate is at the stable value, it means that the on-site personnel have not obtained the earthquake information, so it needs to be prompted in time. If no earthquake data is queried, it is possible that the measured information is incorrect, so an early warning is sent to the security personnel, and the information is recorded, and the security personnel judge how to handle it, which can not only avoid incorrect shutdown, but also convey the actual information for the on-site security personnel to make decisions.

[0161] In a feasible implementation manner, if the device is a power supply device, the method for predicting device faults is as follows:

[0162] Step M1: Collect multi-dimensional operation data of the power supply device in real time.

[0163] Step M1 is specifically: Deploy temperature sensors to monitor the temperature of the power supply heat sink, voltage sensors to monitor the output voltage fluctuation, and vibration sensors to monitor the vibration of internal components. The measurement range of the temperature sensor is -20°C to 120°C, and the accuracy is ±0.5°C; the measurement range of the voltage sensor is 0 to 600V, and the accuracy is ±0.1V; the measurement range of the vibration sensor is 0 to 20mm / s, and the accuracy is ±0.01mm / s.

[0164] Use an industrial Ethernet gateway that supports the Modbus protocol to transmit the collected data in real time.

[0165] Perform preliminary anomaly detection on the temperature data at the edge computing node, and determine it as abnormal when the temperature exceeds 80°C.

[0166] Step M2: Preprocess the collected data and identify abnormal signals.

[0167] Step M2 is specifically: Normalize the temperature data and calculate the temperature change rate as a feature; perform wavelet transform on the voltage data to extract the voltage fluctuation pattern as a feature; calculate the peak factor of the vibration amplitude for the vibration data as a feature.

[0168] Use the WGAN-GP model to generate a balanced data set based on the above features to expand small-sample data;

[0169] The recognition result is obtained by identifying abnormal signals from the balanced data set.

[0170] Step M3: Determine whether there is an abnormal signal according to the recognition result. If so, execute Step M4.

[0171] Step M4: Predict the fault type and transfer path based on the hybrid model (CNN+LSTM) and knowledge graph, and generate control instructions to feedback to the device side.

[0172] Specifically, Step M4 is as follows: Input the preprocessed data into the CNN+LSTM hybrid model for fault prediction. The CNN extracts local features such as voltage fluctuation patterns, and the bidirectional LSTM captures temporal dependencies such as temperature change trends to predict the fault probability within the next 1 hour.

[0173] Correct the prediction result by combining with the mechanism model of the heat dissipation efficiency of the power supply device.

[0174] Based on the knowledge graph, associate historical fault cases such as overload and overheat, and predict the path that the fault may transfer from local overheat to circuit board damage.

[0175] If the fault probability exceeds 90%, generate a control instruction to reduce the output voltage, and generate an instruction to notify the maintenance personnel to check the heat dissipation system.

[0176] Send the control instruction to the power controller for execution through the OPC UA protocol, and at the same time, link with the SCADA (Supervisory Control and Data Acquisition) system to record maintenance information.

[0177] Real-time monitor the operation status of the power supply. If the fault is not resolved, update the model parameters to optimize subsequent predictions.

[0178] In a feasible implementation, if the fault prediction of the production line robot is carried out, the method is as follows:

[0179] Step W1: Real-time collect multi-dimensional operation data of the robot.

[0180] Specifically, Step W1 is as follows: Deploy temperature sensors to monitor the temperature of the servo motor, vibration sensors to monitor the vibration of joint components, and displacement sensors to monitor the displacement of the arm rod. The measurement range of the temperature sensor is 0 to 100 °C, with an accuracy of ±0.2 °C; the measurement range of the vibration sensor is 0 to 25 mm / s, with an accuracy of ±0.05 mm / s; the measurement range of the displacement sensor is 0 to 1000 mm, with an accuracy of ±0.1 mm.

[0181] Use an industrial wireless gateway that supports the OPC UA protocol to transmit the collected data in real time.

[0182] Perform preliminary anomaly detection on the vibration data at the edge computing node, and determine it as an anomaly when the vibration value exceeds 5 mm / s.

[0183] Step W2: Preprocess the collected data and identify abnormal signals.

[0184] Step W2 is specifically as follows: Calculate the temperature rise rate as a feature for the temperature data; Perform wavelet packet decomposition on the vibration data to extract energy features; Calculate the displacement deviation as a feature for the displacement data.

[0185] Use the WGAN-GP model to generate a balanced dataset based on the above features.

[0186] Step W3: Judge whether there is an abnormal signal according to the recognition result. If so, execute Step W4.

[0187] Step W4: Predict the fault type and transfer path based on the hybrid model (CNN+LSTM) and the knowledge graph, and generate control instructions to feedback to the device side.

[0188] Step W4 is specifically as follows: Input the preprocessed data into the CNN+LSTM hybrid model for fault prediction. The CNN extracts local features such as the vibration energy distribution, and the bidirectional LSTM captures the temporal dependencies such as the temperature rise trend to predict the fault probability within the next 4 hours.

[0189] Combine the kinematics and dynamics models of the manipulator to correct the prediction results.

[0190] Based on the knowledge graph, associate historical fault cases such as bearing wear and motor overheating, and predict the path where the fault may transfer from vibration anomaly to displacement deviation.

[0191] If the fault probability exceeds 85%, generate a control instruction to reduce the operating speed, and generate an instruction to notify the maintenance personnel to check the joint components.

[0192] Send the control instructions to the PLC for execution through the industrial Ethernet, and at the same time link with the MES system to record the maintenance information.

[0193] Real-time monitor the operating state of the manipulator. If the fault is not solved, update the model parameters to optimize the subsequent prediction.

[0194] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the device fault prediction method based on the industrial Internet in this application. Based on this technical concept, more forms of simple transformations are within the protection scope of this application.

[0195] This application also provides a device fault prediction device based on the industrial Internet. Please refer to Figure 5 The device fault prediction device based on the industrial Internet includes:

[0196] The acquisition module 10 is used to acquire the target operation data of the device in real time.

[0197] The processing and recognition module 20 is used to preprocess the target operation data and recognize abnormal signals to obtain a recognition result.

[0198] The prediction module 30 is used to, when the recognition result is that an abnormal signal is detected, predict the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result.

[0199] The generation module 40 is used to generate a control instruction based on the prediction result and feedback it to the device side.

[0200] The device for predicting equipment faults based on the industrial Internet provided by this application adopts the method for predicting equipment faults based on the industrial Internet in the above embodiment, and can solve the technical problem of low accuracy of fault prediction. Compared with the prior art, the beneficial effects of the device for predicting equipment faults based on the industrial Internet provided by this application are the same as those of the method for predicting equipment faults based on the industrial Internet provided by the above embodiment, and other technical features in the device for predicting equipment faults based on the industrial Internet are the same as the features disclosed in the method of the above embodiment, which will not be elaborated here.

[0201] In one embodiment, the prediction module 30 is further used to, when the recognition result is that an abnormal signal is detected, input the abnormal operation data into a hybrid model for fault prediction to obtain a fault probability; determine the fault type of the abnormal signal according to the fault probability; associate historical fault cases based on the knowledge graph, and predict the transfer path of the abnormal signal to obtain a predicted fault transfer path; obtain a prediction result through the fault type and the predicted fault transfer path.

[0202] In one embodiment, the prediction module 30 is further used to, when the recognition result is that an abnormal signal is detected, divide the abnormal operation data according to a preset time window to obtain a plurality of subsequences, and the subsequences are expressed as: X = {x1, x2,..., xT}, where X represents a data sequence within a time window, T is the time step, and xt is the operation data at the t-th moment; input the plurality of subsequences into a convolutional neural network in the hybrid model for local feature extraction to obtain local features; splice the local features through a bidirectional LSTM in the hybrid model to obtain spliced features; perform fault prediction on the spliced features through a fully connected layer in the hybrid model to obtain a fault probability within a preset time period.

[0203] In one embodiment, the processing and recognition module 20 is further configured to perform data cleaning, normalization, and feature extraction on the target operation data to obtain preprocessed target operation data; generate an updated data set from the preprocessed target operation data by using a generative adversarial network model; add the updated data set to the preprocessed target operation data to form a balanced data set; and perform abnormal signal recognition on the balanced data set to obtain a recognition result.

[0204] In one embodiment, the processing and recognition module 20 is further configured to reconstruct the balanced data set through a preset autoencoder model to obtain reconstructed output data; calculate a reconstruction error between the balanced data set and the reconstructed output data; compare the reconstruction error with a preset error threshold to perform abnormal signal recognition; and when the reconstruction error is greater than the preset error threshold, determine that the recognition result is that an abnormal signal is detected.

[0205] In one embodiment, the acquisition module 10 is further configured to obtain multi-dimensional operation data in real time through a deployed sensor network, where the multi-dimensional operation data at least includes temperature, voltage, vibration, and noise data; perform preliminary abnormal detection on the multi-dimensional operation data at an edge computing node to obtain a preliminary detection result; and use the multi-dimensional operation data without abnormalities in the preliminary detection result as the target operation data.

[0206] In one embodiment, the generation module 40 is further configured to determine a failure probability, a predicted failure transfer path, and a failure type based on the prediction result; when the failure probability exceeds a preset probability threshold, generate a control instruction according to the failure type and the predicted failure transfer path, and feedback the control instruction to the device side.

[0207] In one embodiment, the generation module 40 is further configured to obtain a vibration amplitude and a noise change rate in the target operation data according to the prediction result; detect Internet earthquake data when the vibration amplitude is the same as the earthquake wave data and the noise change rate is within a preset range; generate a shutdown control instruction when the Internet earthquake data indicates that there is an earthquake; and feedback the shutdown control instruction to the device side so that the device side performs a shutdown process based on the shutdown control instruction after a preset time period.

[0208] This application provides an industrial Internet-based device failure prediction device, where the industrial Internet-based device failure prediction device includes: 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 so that the at least one processor can execute the industrial Internet-based device failure prediction method in the first embodiment above.

[0209] The following refers to Figure 6 , which shows a schematic structural diagram of a device fault prediction device based on the industrial Internet suitable for implementing the embodiments of the present application. The device fault prediction device based on the industrial Internet in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions, tablet computers), PMPs (Portable Media Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The shown device fault prediction device based on the industrial Internet is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0210] As Figure 6 shown, the device fault prediction device based on the industrial Internet may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can execute various appropriate actions and processes according to the program stored in the ROM (Read Only Memory) 1002 or the program loaded from the storage device 1003 into the RAM (Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of the device fault prediction device based on the industrial Internet are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, an LCD (Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the device fault prediction device based on the industrial Internet to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows a device fault prediction device based on the industrial Internet with various systems, it should be understood that it is not required to implement or have all the shown systems. More or fewer systems may be alternatively implemented or had.

[0211] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product that includes a computer program carried on a computer-readable medium, and the computer program contains program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by a processing device 1001, the above-mentioned functions defined in the methods of the embodiments disclosed in the present application are executed.

[0212] The device for predicting equipment failures based on the industrial Internet provided by the present application adopts the method for predicting equipment failures based on the industrial Internet in the above-mentioned embodiments, and can solve the technical problem of low accuracy of failure prediction. Compared with the prior art, the beneficial effects of the device for predicting equipment failures based on the industrial Internet provided by the present application are the same as those of the method for predicting equipment failures based on the industrial Internet provided by the above-mentioned embodiments, and other technical features in the device for predicting equipment failures based on the industrial Internet are the same as the features disclosed in the method of the previous embodiment, and will not be elaborated herein.

[0213] It should be understood that the various parts disclosed in the present application can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0214] As described above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0215] The present application provides a computer-readable storage medium with computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the method for predicting equipment failures based on the industrial Internet in the above-mentioned embodiments.

[0216] The computer-readable storage medium provided by this application can be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or flash memory), optical fibers, CD-ROM (Compact Disk Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, device, or component. The program code contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0217] The above computer-readable storage medium can be included in an industrial Internet-based device fault prediction device; or it can exist independently without being assembled into an industrial Internet-based device fault prediction device.

[0218] The above computer-readable storage medium carries one or more programs. When the one or more programs are executed by an industrial Internet-based device fault prediction device, the industrial Internet-based device fault prediction device is enabled to: real-time collect the target operation data of the device; preprocess the target operation data and identify abnormal signals to obtain an identification result; when the identification result is that an abnormal signal is detected, predict the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result; generate a control instruction based on the prediction result and feedback it to the device end.

[0219] Computer program code for performing the operations of this application can be written in one or more programming languages or combinations thereof. The above-mentioned programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a LAN (Local Area Network) or a WAN (Wide Area Network), or it can be connected to an external computer (for example, by using an Internet service provider to connect through the Internet).

[0220] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of the code, and this module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutively represented blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0221] The modules described in the embodiments of this application can be implemented in software or in hardware. Among them, the name of the module does not constitute a limitation to the unit itself in some cases.

[0222] The readable storage medium provided by this application is a computer-readable storage medium. The computer-readable storage medium stores computer-readable program instructions (i.e., computer programs) for performing the above-mentioned device fault prediction method based on the industrial Internet, and can solve the technical problem of low accuracy of fault prediction. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by this application are the same as those of the device fault prediction method based on the industrial Internet provided by the above embodiments, and will not be elaborated here.

[0223] The present application also provides a computer program product, including a computer program, which when executed by a processor implements the steps of the device fault prediction method based on the industrial Internet as described above.

[0224] The computer program product provided by the present application can solve the technical problem of low accuracy of fault prediction. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the device fault prediction method based on the industrial Internet provided in the above embodiments, and will not be elaborated herein.

[0225] The above are only partial embodiments of the present application, and thus do not limit the patent scope of the present application. Any equivalent structural transformation made under the technical concept of the present application by using the content of the specification and drawings of the present application, or direct / indirect application in other related technical fields, is included in the patent protection scope of the present application.

Claims

1. A device fault prediction method based on the industrial Internet, characterized in that, The device fault prediction method based on the industrial Internet of Things includes: Collecting the target operation data of the device in real time; Preprocessing the target operation data and identifying abnormal signals to obtain an identification result; When the identification result is that an abnormal signal is detected, predicting the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result; Generating a control instruction based on the prediction result and feeding it back to the device side.

2. The method according to claim 1, wherein The step of predicting the fault type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph to obtain a prediction result when the identification result is that an abnormal signal is detected includes: When the identification result is that an abnormal signal is detected, inputting the operation data with anomalies into a hybrid model for fault prediction to obtain a fault probability; Determining the fault type of the abnormal signal according to the fault probability; Associating historical fault cases based on the knowledge graph and predicting the transfer path of the abnormal signal to obtain a predicted fault transfer path; Obtaining a prediction result through the fault type and the predicted fault transfer path.

3. The method according to claim 2, characterized in that The step of inputting the operation data with anomalies into a hybrid model for fault prediction to obtain a fault probability when the identification result is that an abnormal signal is detected includes: When the identification result is that an abnormal signal is detected, dividing the operation data with anomalies according to a preset time window to obtain a plurality of subsequences, and the subsequences are expressed as: X = {x1, x2,..., xT}, where X represents a data sequence within a time window, T is the time step, and xt is the operation data at the t-th moment; Inputting the plurality of subsequences into a convolutional neural network in the hybrid model for local feature extraction to obtain local features; Performing feature splicing on the local features through a bidirectional LSTM in the hybrid model to obtain spliced features; Performing fault prediction on the spliced features through a fully connected layer in the hybrid model to obtain a fault probability within a preset time period.

4. The method according to claim 1, wherein The step of preprocessing the target operation data and identifying abnormal signals to obtain an identification result includes: Performing data cleaning, normalization, and feature extraction on the target operation data to obtain preprocessed target operation data; Generating an updated data set for the preprocessed target operation data by using a generative adversarial network model; Adding the updated data set to the preprocessed target operation data to form a balanced data set; Identifying abnormal signals for the balanced data set to obtain an identification result.

5. The method according to claim 4, wherein The step of identifying abnormal signals for the balanced data set to obtain an identification result includes: Reconstructing the balanced data set through a preset autoencoder model to obtain reconstructed output data; Calculating the reconstruction error between the balanced data set and the reconstructed output data; Comparing the reconstruction error with a preset error threshold for abnormal signal identification; When the reconstruction error is greater than the preset error threshold, determining that the identification result is that an abnormal signal is detected.

6. The method according to claim 1, wherein The step of collecting the target operation data of the device in real time includes: Obtain multi-dimensional operation data in real time through the deployed sensor network, and the multi-dimensional operation data at least includes temperature, voltage, vibration and noise data; Perform preliminary anomaly detection on the multi-dimensional operation data at the edge computing node to obtain a preliminary detection result; Use the multi-dimensional operation data without anomalies in the preliminary detection result as the target operation data.

7. The method according to claim 1, wherein The step of generating a control instruction based on the prediction result and feeding it back to the device side includes: Determine the failure probability, predicted failure transfer path and failure type based on the prediction result; When the failure probability exceeds a preset probability threshold, generate a control instruction according to the failure type and the predicted failure transfer path, and feed the control instruction back to the device side.

8. The method according to claim 1, wherein The step of generating a control instruction based on the prediction result and feeding it back to the device side includes: Obtain the vibration amplitude and noise change rate in the target operation data according to the prediction result; When the vibration amplitude is the same as the seismic wave data and the noise change rate is within a preset range, detect the Internet seismic data; When there is an earthquake in the Internet seismic data, generate a shutdown control instruction; Feed the shutdown control instruction back to the device side so that the device side performs a shutdown process based on the shutdown control instruction after a preset time.

9. An apparatus for predicting equipment failures based on the industrial Internet, characterized in that, The device includes: An acquisition module for real-time acquisition of the target operation data of the device; A processing and recognition module for preprocessing the target operation data and identifying abnormal signals to obtain an identification result; A prediction module for predicting the failure type and transfer path of the abnormal signal based on a hybrid model and a knowledge graph when the identification result is that an abnormal signal is detected, to obtain a prediction result; A generation module for generating a control instruction based on the prediction result and feeding it back to the device side.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the device failure prediction method based on the industrial Internet according to any one of claims 1 to 8 are implemented.

Images