Equipment state prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses an equipment state prediction method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining to-be-predicted equipment operation data; obtaining a state prediction model, wherein the state prediction model is trained through multiple pieces of equipment time sequence data; and performing feature extraction and probability distribution mapping by using the state prediction model according to the equipment operation data, and outputting an equipment prediction state. The device state prediction method and device can realize device state prediction, improve the efficiency and accuracy of device state prediction, help to predict the working state and fault possibility of the device in advance, avoid heavy loss caused by failure to discover the device operation fault in time, and can be widely applied to the technical field of industrial devices.

Description

Technical Field

[0001] The present application relates to the technical field of industrial equipment, and in particular to a method, device, electronic device and storage medium for predicting equipment status. Background Art

[0002] As a core technical means in modern industrial maintenance management, online monitoring and predictive maintenance of industrial equipment plays an irreplaceable role in preventing similar major safety accidents. By collecting, analyzing and interpreting relevant data under the operating status of the equipment online, early signals of potential failures inside the equipment can be captured in time. Therefore, it is particularly necessary to predict the status of industrial equipment, which can help staff predict the working status and failure possibility of the equipment in advance, so that staff can provide more timely responses. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a device state prediction method, device, electronic device and storage medium, which can realize device state prediction and improve the accuracy of device state prediction.

[0004] On the one hand, an embodiment of the present application proposes a device state prediction method, the method comprising the following steps:

[0005] Obtain the equipment operation data to be predicted;

[0006] Acquire a state prediction model, wherein the state prediction model is trained by time series data of multiple devices;

[0007] The state prediction model is used to perform feature extraction and probability distribution mapping according to the equipment operation data, and the predicted state of the equipment is output.

[0008] In some embodiments, the acquiring state prediction model specifically includes:

[0009] Get a sample dataset;

[0010] A state prediction model is constructed, and the state prediction model is trained and evaluated using the sample data set.

[0011] In some embodiments, obtaining the sample data set specifically includes:

[0012] Collect device time series data corresponding to multiple prediction devices;

[0013] Performing data normalization processing on the time series data of each device to obtain normalized time series data of each device;

[0014] Get multiple device status prediction labels;

[0015] Data transformation is performed on each of the normalized device time series data, and a target device prediction state label corresponding to each of the device time series data is determined from a plurality of the device state prediction labels.

[0016] In some embodiments, the constructing of the state prediction model, using the sample data set to train and evaluate the state prediction model, specifically includes:

[0017] Dividing the sample data set into a training data set and a validation data set;

[0018] Using the training data set to train the state prediction model, using the verification data set to perform model evaluation on the state prediction model, and determining a model evaluation result;

[0019] Determine whether to continue training the state prediction model based on the model evaluation result.

[0020] In some embodiments, the state prediction model includes a first feature extraction module, a second feature extraction module and a feature mapping module, and the state prediction model is used to perform feature extraction and probability distribution mapping according to the device operation data, and output the device prediction state, specifically including:

[0021] Inputting the equipment operation data into the state prediction model, performing feature extraction and convolution operations on the equipment operation data using the first feature extraction module, and obtaining first feature data corresponding to the equipment operation data;

[0022] Using the second feature extraction module to perform word embedding conversion and feature fusion on the device operation data to obtain corresponding second feature data, and performing feature fusion according to the first feature data and the second feature data to obtain corresponding feature fusion data;

[0023] The feature mapping module is used to perform probability distribution mapping on the feature fusion data, and the device state prediction label with the highest probability is determined from the multiple device state prediction labels, and the device state prediction label with the highest probability is obtained as the device prediction state.

[0024] In some embodiments, the method further comprises:

[0025] Acquire feature weight data, and generate a corresponding class activation map according to the feature weight data, wherein the feature weight data is obtained when the second feature extraction module performs feature fusion on the first feature data and the second feature data.

[0026] In some embodiments, the acquiring feature weight data and generating a corresponding class activation map according to the feature weight data specifically includes:

[0027] Obtaining a data matrix corresponding to the equipment operation data;

[0028] The feature weight data is superimposed and mapped to the data matrix to generate the class activation map.

[0029] On the other hand, an embodiment of the present application provides a device for predicting a device state, the device comprising:

[0030] The first module is used to obtain the equipment operation data to be predicted;

[0031] The second module is used to obtain a state prediction model, wherein the state prediction model is trained by time series data of multiple devices;

[0032] The third module is used to use the state prediction model to perform feature extraction and probability distribution mapping according to the equipment operation data, and output the predicted equipment state.

[0033] On the other hand, an embodiment of the present application provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the device state prediction method described above when executing the computer program.

[0034] On the other hand, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the device state prediction method described above is implemented.

[0035] The embodiments of the present application include at least the following beneficial effects: the present application provides a device state prediction method, device, electronic device and storage medium, which obtains the device operation data to be predicted, obtains the state prediction model, uses the state prediction model to perform feature extraction and probability distribution mapping based on the device operation data, and outputs the predicted device state. The present application can realize device state prediction, improve the efficiency and accuracy of device state prediction, help predict the working state and possible failure of the device in advance, and avoid significant losses caused by failure to timely discover device operation failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of a device status prediction method provided by an embodiment of the present application;

[0037] Figure 2 It is a schematic diagram of a class activation diagram in an embodiment of the present application;

[0038] Figure 3 It is a structural schematic diagram of a device state prediction apparatus provided in an embodiment of the present application;

[0039] Figure 4 It is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0041] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0042] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0044] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0045] Before describing the embodiments of the present application in detail, some nouns and terms involved in the embodiments of the present application are first described. The nouns and terms involved in the embodiments of the present application are subject to the following explanations.

[0046] 1) Process industry: It is an industrial sector that produces goods or semi-finished products through the transformation of matter or energy based on continuous or batch processes such as chemical, physical or biological processes. This type of industry usually deals with raw materials, which are transformed into final products through a series of processes, and these processes are often irreversible. The typical characteristics of process industries include continuity, scale and high automation.

[0047] 2) Digital twin: refers to a technical concept that builds a virtual model of a physical entity or system in a digital environment to monitor, analyze and optimize its entire life cycle. Digital twin technology uses sensors, IoT devices, data analysis and machine learning to map the real-time status, behavior and performance of a physical entity to its virtual copy, enabling real-time monitoring, simulation prediction, fault diagnosis and remote control.

[0048] 3) Transformer: It is a deep learning model architecture. Its core innovation lies in its attention-based architecture, which abandons the traditional recurrent neural network (RNN) and convolutional neural network (CNN) structures, thereby achieving more efficient and flexible model training and reasoning.

[0049] 4) Softmax: It is a commonly used activation function, which is widely used in multi-classification machine learning models, especially in the output layer of neural networks. It converts a real vector into a probability distribution so that each output value is between 0 and 1, and the sum of all output values ​​is 1.

[0050] Digital twin is an advanced modeling technology that creates a virtual replica of a physical entity to simulate, predict and optimize the entity's behavior and performance in real time. With the advancement of new-generation electronic information technologies such as the Internet of Things, big data, cloud computing, and artificial intelligence, the implementation of digital twins has gradually become possible. In addition to the aerospace field, digital twins are also used in industries such as electricity, shipping, manufacturing, and oil and gas. Especially in the field of industrial processes, digital twins are considered to be an effective means to achieve interactive integration between the manufacturing information world and the physical world.

[0051] In the field of industrial equipment technology, existing digital twin technology has the following shortcomings:

[0052] 1) When digital twin modeling of industrial equipment is carried out in a data-driven manner, deep learning technology needs to be used to learn and analyze a large amount of historical data of process industrial equipment. The on-site system of process industrial equipment is basically in a normal operating state, so the data generated by the equipment are all normal data. Therefore, the deep learning network model can only learn the data mode of industrial equipment in a normal state, lacking data in an abnormal state. The data set has a serious imbalance problem, which will cause the deep learning model to be unable to predict the abnormal state of the equipment and the prediction effect is poor. In the prior art, in order to obtain abnormal data of industrial equipment, it is necessary to use the mechanism model of industrial equipment to simulate the abnormal state. However, the equipment and mechanism reactions involved in the industrial process are complex and numerous. The cost of using the mechanism model for simulation is too high, and it is difficult to cover all industrial equipment.

[0053] 2) Existing deep learning network models based on data-driven digital twin technology mostly focus on local feature extraction, but relatively few designs on multi-source heterogeneous time-space feature extraction. They lack multi-dimensional, multi-modal, and multi-feature expressions, and the information expression is not clear enough, which limits the design ideas of robust and efficient digital twin modeling.

[0054] 3) The data-driven process industry digital twin model is constructed using deep learning technology. Although the deep learning network model can learn historical data, predict future data, and obtain prediction results, it is impossible to derive the basis for the deep learning model to judge the prediction results. Therefore, the data-driven process industry digital twin model lacks a certain degree of interpretability.

[0055] Based on this, the embodiments of the present application propose a device state prediction method, device, electronic device and storage medium, which can use a deep learning network model to divide the device state into different category labels according to the fluctuation amplitude of the equipment operation data, improve the quality of the historical data set, enable the digital twin model to better learn the data, and dig out the basis for the deep learning network model to make predictions and judgments, visualize the data that contributes the most to the prediction, provide a certain degree of interpretability for the digital twin model, enhance the modeling ability of the digital twin model, and effectively improve the accuracy of the data-driven digital twin model in predicting the state of industrial equipment.

[0056] Reference Figure 1 , Figure 1 This is an optional flowchart of a device state prediction method provided in an embodiment of the present application. The method may include but is not limited to steps S101 to S103:

[0057] Step S101, obtaining the equipment operation data to be predicted;

[0058] Step S102, obtaining a state prediction model, where the state prediction model is trained using time series data of multiple devices;

[0059] Step S103, using the state prediction model to perform feature extraction and probability distribution mapping based on the equipment operation data, and output the predicted state of the equipment.

[0060] In some embodiments, the state prediction model is a deep learning network model. Optionally, multi-source heterogeneous equipment operation data is input into the state prediction model, and the state prediction model is used to extract the time-space features of the multi-source heterogeneous data, output the equipment prediction state, and improve the accuracy of the equipment state prediction.

[0061] In some embodiments, step S102 may include but is not limited to steps S201 to S202:

[0062] Step S201, obtaining a sample data set;

[0063] Step S202: construct a state prediction model, and use the sample data set to train and evaluate the state prediction model.

[0064] In some embodiments, step S201 may include but is not limited to steps S301 to S304:

[0065] Step S301, collecting device time series data corresponding to multiple prediction devices;

[0066] Step S302, performing data normalization processing on the time series data of each device to obtain normalized time series data of each device;

[0067] Step S303, obtaining multiple device status prediction tags;

[0068] Step S304, performing data transformation on the normalized time series data of each device, and determining the target device prediction state label corresponding to each device time series data from a plurality of device state prediction labels.

[0069] In some embodiments, the prediction device is a process industry device, and the device time series data corresponding to multiple prediction devices are collected, and the prediction device is defined as X 1 , X 2 , …, X n , where the predicted device X n The equipment time series data generated within a specified period of time can be expressed as {x M 1,x M 2,…,x M t}, where M∈[1,n], x M tIt represents the device data of the Mth predicted device at time t, that is, each predicted device has corresponding time series data, and n is the number of predicted devices.

[0070] Then, the time series data of each device is normalized. Optionally, the Min-Max (minimum value - maximum value) normalization method is used for normalization. The calculation formula is as follows:

[0071]

[0072] Among them, x norm is the normalized data of x, where x refers to the time series data of the Mth device {x M 1,x M 2,…,x M t}, x max is the time series data of the Mth device {x M 1,x M 2,…,x M t}, x min The Mth device time series data {x M 1,x M 2,…,x M t} is the minimum value among them.

[0073] After normalizing the device time series data, the data features can be scaled to a specific range, usually between 0 and 1, to eliminate the impact of different dimensions between different device data on the classification model, which helps to improve the training efficiency and performance of the deep learning network model.

[0074] For example, since the data values ​​of various process industrial equipment fluctuate in different amplitudes, in order to make the outlier feature points with larger fluctuations more prominent as much as possible and reduce the interference of equipment data noise fluctuations on the model, the normalized equipment time series data corresponding to a certain prediction equipment is transformed by the following formula:

[0075]

[0076] Among them, x′ is the target device prediction state label of the normalized device time series data, is the mean value of the normalized device time series data, std is the standard deviation of the normalized device time series data, thr is the custom classification threshold, and 0, 1, 2, and 3 are all different device prediction status labels.

[0077] After the above data transformation processing, the target device prediction state label corresponding to each prediction device can be obtained as the classification prediction task target of the deep learning neural network, that is, predicting the target device prediction state of the prediction device at the next moment.

[0078] In some embodiments, step S202 may include but is not limited to steps S401 to S403:

[0079] Step S401, dividing the sample data set into a training data set and a verification data set;

[0080] Step S402, training the state prediction model using the training data set, performing model evaluation on the state prediction model using the verification data set, and determining the model evaluation result;

[0081] Step S403: Determine whether to continue training the state prediction model based on the model evaluation result.

[0082] In some embodiments, the sample data set is divided into a training data set and a validation data set in a certain ratio (eg, 4:1) for training and model evaluation of the state prediction model.

[0083] Optionally, during the model training process, a cross entropy loss function is used as a training loss function. The cross entropy loss function is as follows:

[0084]

[0085] Among them, L is the cross entropy loss value, N is the number of sample training data, C is the number of device status categories, and y ij is the true label of sample training data i in device state category j. The cross entropy loss is calculated based on the prediction result and the true label, and then the gradient is calculated by the chain rule to update the neural network weights of the state prediction model. The iterative optimizer used in model training is the Adam optimizer, and the learning rate can usually be set to 0.001. The parameters can be adjusted according to the actual performance. After repeated iterative training optimization until the loss converges or reaches the preset number of training rounds, the trained state prediction model is obtained.

[0086] Then, the generalization ability of the state prediction model on unseen data is evaluated through the verification data set. The model evaluation indicators used include accuracy, precision, recall, F1 score, etc., and the specific value corresponding to each model evaluation indicator is determined. When the model evaluation result shows that the specific value corresponding to each model evaluation indicator exceeds the threshold set for each model evaluation indicator, the training of the state prediction model is terminated, otherwise, the state prediction model continues to be trained.

[0087] In some embodiments, step S103 may include but is not limited to steps S501 to S503:

[0088] Step S501, inputting the equipment operation data into the state prediction model, using the first feature extraction module to perform feature extraction and convolution operation on the equipment operation data, and obtaining first feature data corresponding to the equipment operation data;

[0089] Step S502, using a second feature extraction module to perform word embedding conversion and feature fusion on the device operation data to obtain corresponding second feature data, and performing feature fusion based on the first feature data and the second feature data to obtain corresponding feature fusion data;

[0090] Step S503, using a feature mapping module to perform probability distribution mapping on the feature fusion data, determining the device state prediction label with the highest probability from multiple device state prediction labels, and obtaining the device state prediction label with the highest probability as the device prediction state.

[0091] In some embodiments, the device operation data is device time series data of the target device, and the first feature extraction module is used to perform feature extraction and convolution operations on the device operation data. Specifically, the first feature extraction module first uses a one-dimensional linear transformation to perform time series feature extraction on the device operation data of the target device, and maps the original input data to a higher-dimensional feature space. The formula is as follows:

[0092] z=xA T +b;

[0093] Wherein, z is the first feature data obtained after feature extraction of the equipment operation data, x is the equipment operation data, A T is the weight matrix of the state prediction model, and b is the bias coefficient.

[0094] Then, a two-dimensional convolution kernel is used to perform three-layer convolution operations on the first feature data. Then, the ReLU activation function is used to introduce nonlinearity after each layer of convolution. At the same time, the maximum pooling layer is used to compress the extracted feature information. After each convolution and maximum pooling layer, the feature dimension will be halved, and the local features in the time and space dimensions (such as the first feature data mentioned above) can be gradually extracted. Finally, average pooling is used to map the local features to the same dimension.

[0095] In some embodiments, the second feature extraction module is designed based on the Transformer model, and the second feature extraction module is used to perform word embedding conversion and feature fusion on the device operation data. The second feature extraction module includes a word embedding unit, a position encoding unit and a multi-head attention mechanism unit. Optionally, the device operation data is first input into the word embedding unit, converted into a word embedding vector, and the word embedding vector is divided into blocks at the same time, and the divided word embedding vector is input into the position encoding unit. Since the Transformer model does not have the built-in ability to process sequence order, position encoding is introduced to provide position information. The position encoding generated by the sine and cosine functions is shown in the following formula:

[0096]

[0097] Among them, x refers to the time series data of the Mth device {x M 1,x M 2,…,x M t}, pos is the position of the word embedding vector after the data X is divided, i is the dimension index, d model is the dimension of the model, and PE() is the position encoding function.

[0098] Then, the position-encoded data is input into the encoder, which is composed of N identical encoder layers stacked together. Each encoder layer includes a multi-head mutual attention mechanism, a feedforward fully connected network, a residual connection, and layer normalization to extract the global key features of the spatiotemporal data of the equipment operation data (such as the second feature data mentioned above).

[0099] A multi-head attention mechanism unit is used to fuse the local features extracted by the above model with the global key features of spatiotemporal data. The multi-head attention mechanism can capture the different degrees of relationship between input features through each head respectively, and can also flexibly adjust the weights between different heads. It has very good adaptability and flexibility, thereby improving the model's ability to model complex relationships, fuse features, and obtain feature fusion data.

[0100] In some embodiments, the feature mapping module includes a normalization layer, a fully connected layer, and a Softmax activation function. Optionally, after the above-mentioned feature fusion data is input into the normalization layer, the fully connected layer, and the Softmax activation function in sequence, the output can be mapped to the probability distribution of each device prediction state label. The device prediction state label with the highest probability is the device prediction state predicted by the model.

[0101] In some embodiments, the above device status prediction method may further include step S601:

[0102] Step S601, acquiring feature weight data, and generating a corresponding class activation map according to the feature weight data, wherein the feature weight data is obtained when a second feature extraction module performs feature fusion on the first feature data and the second feature data.

[0103] In step S601 of some embodiments, the class activation map is a heat map used to show which areas in the input image contribute most to the prediction of a specific category. It highlights the area that the model "focuses on" by superimposing color or brightness on the image, obtains the data matrix corresponding to the device operation data, and superimposes and maps the feature weight data to the data matrix to generate a class activation map.

[0104] Specifically, the class activation map technology is used to extract the gradient weight matrix of the state prediction model in the prediction process. The gradient weight matrix is ​​a feature weight matrix output by the multi-head attention mechanism in the prediction process of the above model. The specific process is as follows:

[0105] First, calculate the prediction score y of the device prediction state c c Relative to the attention weight matrix A of the hth attention head in the multi-head attention mechanism unit h The gradient is: Then for each attention head h, the gradients of all attention weight matrices are averaged to obtain the weight a of the attention head h , the calculation formula is as follows:

[0106]

[0107] Among them, Z is the size of the attention weight matrix A. Then, the gradient weight matrix L is calculated by weighted summing the weights of each attention head. The calculation formula is as follows:

[0108]

[0109] Among them, ReLU is a commonly used activation function that sets negative input values ​​to 0 and keeps positive values ​​unchanged.

[0110] Finally, the gradient weight matrix L is superimposed on the data matrix and mapped to the degree of correlation of the industrial equipment data, thereby reflecting the contribution of each data value in the equipment operation data to the model prediction results, improving the transparency of the model, and giving the digital twin model a certain degree of interpretability.

[0111] In some embodiments, illustratively, taking the prediction of the state of water treatment related equipment as an example, the above-mentioned equipment state prediction method is applied to predict the equipment state of water treatment related equipment, as follows:

[0112] The first step is to collect the equipment operation data of the target equipment. The equipment operation data records the status information of 16 sensors in the target equipment, including flow, pressure, liquid level, pump status, valve status, etc. The equipment operation data records the operation status data of each sensor in the equipment at different time points, including timestamp, sensor data, sensor identifier and actuator status. The sampling period of each operation status data is 1 second.

[0113] In the second step, the operating status data of each sensor from T+1 to T+16 seconds is used as data input to form a 16*16 matrix data. The state prediction model is used to predict the device prediction state corresponding to the target device at T+17 seconds. The matrix data is transformed into a one-dimensional linear transformation, and the 16*16 matrix data is transformed into a feature vector with a length of 128 to enhance the feature representation and help the model capture the complex relationships in the historical data of the target device.

[0114] The third step is to perform local feature extraction on the 16*16 matrix data, perform three-layer convolution operations on the input data in sequence, and activate it through the ReLU activation function after each convolution. Then, use the maximum pooling layer to reduce the dimension of the features, extract higher-level feature representation vectors, gradually extract local key features, and finally use the average pooling layer to represent the features as the first feature vector of length 128. Then, perform global feature extraction on the 16*16 matrix data, first perform block word embedding on the input data, then add position encoding information to the word embedding vector, and then input it into the Transformer encoder, and also extract the second feature vector of length 128.

[0115] The fourth step is to input the extracted first feature vector and second feature vector into the feature fusion module, use the multi-head attention mechanism to capture the different degrees of relationship between the input features, flexibly adjust the weights between different features, and fuse and output the third feature vector of length 128 (such as the above feature fusion data);

[0116] In the fifth step, the third eigenvector is normalized, and then passed through the fully connected layer and Softmax activation, and finally the final device prediction status is output after multiple iterations of training;

[0117] The sixth step is to use the class activation map technology to obtain the gradient weight matrix in the state prediction model, and superimpose the gradient weight matrix on the data matrix corresponding to the equipment operation data, such as Figure 2As shown in the class activation map obtained after superposition, the warm area (such as the data in the red box) is the data area that the prediction model pays more attention to, and it is also the data area that contributes the most to the prediction. The cold area (such as the data in the green box) is the data area that the prediction model pays less attention to, and it is the data area that contributes the least to the prediction. This improves the transparency of the prediction model and gives the model a certain degree of interpretability.

[0118] Reference Figure 3 , Figure 3 : is an optional structural diagram of a device state prediction apparatus provided in an embodiment of the present application. The apparatus can be used to implement the above-mentioned device state prediction method. The apparatus may include:

[0119] The first module is used to obtain the equipment operation data to be predicted;

[0120] The second module is used to obtain a state prediction model, which is trained by time series data of multiple devices;

[0121] The third module is used to use the state prediction model to perform feature extraction and probability distribution mapping according to the equipment operation data, and output the predicted state of the equipment.

[0122] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0123] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above device state prediction method when executing the computer program. The electronic device can be any smart terminal including a tablet computer.

[0124] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0125] See also Figure 4 , Figure 4 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0126] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0127] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the device state prediction method of the embodiment of this application;

[0128] Input / output interface 903, used to implement information input and output;

[0129] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0130] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0131] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0132] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned device state prediction method is implemented.

[0133] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0134] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0135] The embodiments of the present application provide a device state prediction method, device, electronic device, and storage medium, which obtains the device operation data to be predicted, obtains the state prediction model, uses the state prediction model to perform feature extraction and probability distribution mapping based on the device operation data, and outputs the predicted state of the device. The embodiments of the present application can realize device state prediction, improve the efficiency and accuracy of device state prediction, help predict the working state and possible failure of the device in advance, and avoid significant losses caused by failure to timely discover device operation failures.

[0136] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0137] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0138] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on 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.

[0139] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0140] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0141] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0142] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0143] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0144] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0145] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.

[0146] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0147] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A device state prediction method, characterized in that: The method comprises the following steps: Obtain the equipment operation data to be predicted; Acquire a state prediction model, wherein the state prediction model is trained by time series data of multiple devices; The state prediction model is used to perform feature extraction and probability distribution mapping according to the equipment operation data, and the predicted state of the equipment is output.

2. The device status prediction method according to claim 1, characterized in that: The acquisition state prediction model specifically includes: Get a sample dataset; A state prediction model is constructed, and the state prediction model is trained and evaluated using the sample data set.

3. The device status prediction method according to claim 2, characterized in that: The obtaining of the sample data set specifically includes: Collect device time series data corresponding to multiple prediction devices; Performing data normalization processing on the time series data of each device to obtain normalized time series data of each device; Get multiple device status prediction labels; Data transformation is performed on each of the normalized device time series data, and a target device prediction state label corresponding to each of the device time series data is determined from a plurality of the device state prediction labels.

4. The device status prediction method according to claim 2, characterized in that: The constructing of the state prediction model and using the sample data set to train and evaluate the state prediction model specifically includes: Dividing the sample data set into a training data set and a validation data set; Using the training data set to train the state prediction model, using the verification data set to perform model evaluation on the state prediction model, and determining a model evaluation result; Determine whether to continue training the state prediction model based on the model evaluation result.

5. The device status prediction method according to claim 3, characterized in that: The state prediction model includes a first feature extraction module, a second feature extraction module and a feature mapping module. The state prediction model is used to perform feature extraction and probability distribution mapping according to the device operation data, and output the predicted state of the device, specifically including: Inputting the equipment operation data into the state prediction model, performing feature extraction and convolution operations on the equipment operation data using the first feature extraction module, and obtaining first feature data corresponding to the equipment operation data; Using the second feature extraction module to perform word embedding conversion and feature fusion on the device operation data to obtain corresponding second feature data, and performing feature fusion according to the first feature data and the second feature data to obtain corresponding feature fusion data; The feature mapping module is used to perform probability distribution mapping on the feature fusion data, and the device state prediction label with the highest probability is determined from the multiple device state prediction labels, and the device state prediction label with the highest probability is obtained as the device prediction state.

6. The device state prediction method according to claim 5, characterized in that: The method further comprises: Acquire feature weight data, and generate a corresponding class activation map according to the feature weight data, wherein the feature weight data is obtained when the second feature extraction module performs feature fusion on the first feature data and the second feature data.

7. The device state prediction method according to claim 6, characterized in that: The acquiring feature weight data and generating a corresponding class activation map according to the feature weight data specifically includes: Obtaining a data matrix corresponding to the equipment operation data; The feature weight data is superimposed and mapped to the data matrix to generate the class activation map.

8. A device for predicting equipment status, characterized in that: The device comprises: The first module is used to obtain the equipment operation data to be predicted; The second module is used to obtain a state prediction model, wherein the state prediction model is trained by time series data of multiple devices; The third module is used to use the state prediction model to perform feature extraction and probability distribution mapping according to the equipment operation data, and output the predicted equipment state.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the device state prediction method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the device state prediction method according to any one of claims 1 to 7 is implemented.

Images