Industrial time series data anomaly detection method and system based on data enhancement

Abstract

The invention relates to an industrial time series data anomaly detection method and system based on data enhancement. The method comprises the steps of collecting industrial time series data in an industrial production process; performing data enhancement on the industrial time series data in combination with a data interpolation method and a data preprocessing method; and carrying out anomaly detection on the industrial time series data after data enhancement by adopting a 1D single-layer convolutional neural network model added with a time attention mechanism. According to the method, interpolation enhancement is carried out on the original industrial time series data, the data sample capacity is improved, meanwhile, the data sampling frequency is improved, and the method plays a key role in restoring key information contained in real data for the enhanced data. According to the method, time sequence data input is processed by adopting the 1D convolutional neural network, abnormal conditions are further detected by adopting a softmax classification method through feature extraction, and a time sequence problem is converted into graphic features, so that the detection efficiency is improved, and the method is simple and convenient to use.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for detecting anomalies in industrial time series data based on data enhancement. Background Art

[0002] Anomaly detection refers to the automatic identification of abnormal phenomena embedded in a large amount of normal data. In the industrial manufacturing field, the production process is accompanied by a large amount of production data. With the vigorous development of sensor technology and Internet of Things technology, it has become possible to identify and judge abnormal patterns that do not conform to expected states by mining and analyzing data.

[0003] The task goal of anomaly detection is to determine which data has characteristic patterns that are significantly different from other normal data. Usually, considering that the outliers deviate far from the average of other values, the traditional method uses data statistics to detect deviations to determine whether an anomaly occurs. In the field of industrial manufacturing, with the emergence of the Internet of Things (IoT), a large amount of time series data that integrates different data domains such as multidimensional, network flow, local flow, and hybrid flow has been generated. Therefore, the analysis and anomaly detection methods of time series data in these industrial production processes are related to production safety, which is very important and worthy of study.

[0004] Existing anomaly detection schemes mainly use traditional statistical analysis methods, such as statistical process control (SPC) chart methods such as cumulative sum charts to detect changes in the characteristics of the basic distribution of data to identify anomalies, which requires high accuracy in the original data collection.

[0005] Existing anomaly detection solutions have a strong dependence on the quality of raw data. However, in the industrial production process, there are many uncertainties in the process of industrial Internet of Things collecting and uploading production data. The collection frequency, data transmission rate, data noise quality, etc. will have a great impact on the statistical stability of the results, resulting in poor robustness and weak stability of existing anomaly detection solutions. Summary of the invention

[0006] In view of this, it is necessary to provide an industrial time series data anomaly detection method based on data enhancement to solve the above-mentioned defects of the prior art.

[0007] In order to solve the above problems, in a first aspect, an embodiment of the present invention provides an industrial time series data anomaly detection method based on data enhancement, comprising:

[0008] Collect industrial time series data in industrial production processes;

[0009] Combining a data interpolation method and a data preprocessing method, the industrial time series data is enhanced;

[0010] A 1D single-layer convolutional neural network model with a temporal attention mechanism is used to perform anomaly detection on industrial time series data after data enhancement.

[0011] The process of collecting industrial time series data in the industrial production process also includes:

[0012] Based on Shannon sampling theorem and prior knowledge, the optimal sampling frequency of industrial time series data is determined; wherein the industrial time series data includes voltage, current, temperature, humidity and speed.

[0013] Preferably, the data interpolation method and the data preprocessing method are combined to perform data enhancement on the industrial time series data, including:

[0014] Using the data interpolation method, the industrial time series data X with a length of N is T Expand to length N * Intermediate data Wherein, the data interpolation method is linear interpolation, cubic interpolation or Lagrange interpolation;

[0015]

[0016] In the formula, Indicates the data interpolation method;

[0017] Intermediate data generated by interpolation Data preprocessing is performed to obtain industrial time series data after data enhancement; wherein the data preprocessing includes data rotation, data jitter, data scaling and data rearrangement.

[0018] Preferably, the intermediate data generated by interpolation Perform data preprocessing to obtain industrial time series data after data enhancement, including:

[0019] Assume that the data before time series data preprocessing is The preprocessed data is X' T ;

[0020] Intermediate data generated by data interpolation Four data operations are performed: data rotation, data jitter, data scaling, and data rearrangement. Each operation has two possibilities: use or not use, for a total of 16 possibilities. For each data t∈[1,N * ], generates a random integer operator Ind(i) = random{0,1} of 0 or 1, corresponding to the usage of each data operation, through the matrix To indicate which types of data operations are taken in the preprocessing operation;

[0021] The data preprocessing operation is described as:

[0022]

[0023] In the formula, Represents data rotation operation, Indicates data jitter operation, represents a data scaling operation, Represents a data shuffle operation.

[0024] Preferably, the 1D single-layer convolutional neural network model with a temporal attention mechanism is used to perform anomaly detection on the industrial time series data after data enhancement, including:

[0025] The strategy of aggregating labels of adjacent points is adopted to migrate the label information of the original dataset to the enhanced dataset by calculating the time domain aggregation feature index.

[0026] Using the migrated label information and the temporal attention mechanism, a 1D single-layer convolutional neural network model is trained.

[0027] The time series data is converted into two-dimensional data through the trained 1D single-layer convolutional neural network model. The features are extracted through the convolution layer, and then the distribution position of the abnormal data is output after maximum pooling and softmax classification.

[0028] Preferably, the strategy of adopting adjacent point label aggregation to migrate the label information of the original data set to the enhanced data set by calculating the time domain aggregation feature index includes:

[0029] For the data element x in the original industrial time series dataset i , data element x i The time domain aggregation feature index during training is expressed as:

[0030]

[0031] Among them, N i is the neighborhood of node i in industrial time series data, x j is the neighborhood value, LeakyReLU(·) is a nonlinear activation function used in the training process, a ij is the aggregate score between data elements i and j;

[0032] Neighborhood normalization is performed through the softmax function to calculate the aggregation score of each neighborhood point;

[0033]

[0034] Among them, d(x i ,x j ) is the data element x i ,xj The distance between

[0035] Based on the aggregated scores, the labels in the original data are transferred to the enhanced data points.

[0036] In a second aspect, an embodiment of the present invention provides an industrial time series data anomaly detection system based on data enhancement, comprising:

[0037] Data acquisition module, used to collect industrial time series data in the industrial production process;

[0038] A data enhancement module, used to enhance the industrial time series data by combining a data interpolation method and a data preprocessing method;

[0039] The anomaly detection module is used to detect anomalies in industrial time series data after data enhancement by using a 1D single-layer convolutional neural network model with a temporal attention mechanism.

[0040] In a third aspect, the present invention further provides an electronic device, including a memory and a processor, wherein:

[0041] The memory is used to store programs;

[0042] The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the industrial time series data anomaly detection method based on data enhancement as described in the embodiment of the first aspect of the present invention.

[0043] In a fourth aspect, the present invention further provides a computer-readable storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the industrial time series data anomaly detection method based on data enhancement as described in the embodiment of the first aspect of the present invention.

[0044] Compared with the prior art, the industrial time series data anomaly detection method and system based on data enhancement provided by the present invention have the following beneficial effects:

[0045] 1) The present invention increases the data sample capacity and the data sampling frequency by interpolating and enhancing the original industrial time series data, which plays a key role in restoring the key information contained in the real data from the enhanced data.

[0046] 2) The present invention combines data interpolation methods and data preprocessing methods to perform data enhancement on industrial time series data, thereby improving the richness and diversity of data samples, which is beneficial to parameter convergence and anti-overfitting in subsequent deep learning-based model training.

[0047] 3) In the process of model training, the present invention introduces the strategy of time-domain aggregation feature index, which not only preliminarily extracts the temporal features of the data in the training data set, but also realizes the automatic labeling of the enhanced data, reduces the labor cost and time cost of manual labeling, and improves the efficiency of model training;

[0048] 4) The present invention uses a 1D convolutional neural network to process time series data input, and further uses a softmax classification method to detect abnormal situations through feature extraction, converting time series problems into graphic features, thereby improving detection efficiency and being simple and convenient to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flow chart of the industrial time series data anomaly detection method based on data enhancement provided by the present invention;

[0050] Figure 2 A structural diagram of a 1D single-layer convolutional neural network model provided by the present invention;

[0051] Figure 3 A structural block diagram of the industrial time series data anomaly detection system based on data enhancement provided by the present invention;

[0052] Figure 4 This is a structural block diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0053] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0054] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0055] At present, the existing anomaly detection schemes mainly use traditional statistical analysis methods, such as statistical process control (SPC) chart methods such as cumulative sum charts to detect changes in the characteristics of the basic distribution of data to identify anomalies, which requires high accuracy in the original data collection.

[0056] Existing anomaly detection solutions have a strong dependence on the quality of raw data. However, in the industrial production process, there are many uncertainties in the process of industrial Internet of Things collecting and uploading production data. The collection frequency, data transmission rate, data noise quality, etc. will have a great impact on the statistical stability of the results, resulting in poor robustness and weak stability of existing anomaly detection solutions.

[0057] In view of this, the present invention provides an industrial time series data anomaly detection method based on data enhancement. For massive time series data from any single source, the original data is enhanced and then anomaly detection is performed using a deep neural network model, effectively overcoming the detection difficulties caused by data quality in existing methods. The following will be explained and introduced through multiple embodiments.

[0058] Figure 1 The flowchart of the method for detecting anomalies of industrial time series data based on data enhancement provided by the present invention is as follows. Figure 1 As shown, the industrial time series data anomaly detection method based on data enhancement includes steps S1 to S3, wherein:

[0059] Step S1, collecting industrial time series data in the industrial production process;

[0060] Specifically, in the industrial production process, sensors deployed on equipment collect time series data of detection indicators such as voltage, current, temperature, humidity, speed, etc. These data are usually recorded in discrete form, and each data point contains a timestamp and the corresponding measurement value. T , which can be defined as

[0061] X T ={x1, x2, ..., x i}

[0062] Where, X T Represents the measured value of the detection index X at time T, x i is the measurement value of node i in the industrial time series data, i∈[1, 2, …, T].

[0063] In the process of collecting industrial time series data in the industrial production process, the optimal sampling frequency of the industrial time series data is determined based on Shannon sampling theorem and prior knowledge; wherein the industrial time series data includes voltage, current, temperature, humidity and speed.

[0064] It is understandable that Shannon's sampling theorem states that in order to ensure the complete recovery of the original continuous-time signal, the sampling frequency must be at least twice the highest frequency of the signal. In industrial production, the highest frequency of the signal is usually determined by equipment characteristics and process requirements.

[0065] In this embodiment, the highest frequency fH of the signal can be obtained based on prior knowledge, that is, the highest frequency of the signal can be estimated based on the historical data of the equipment or process and expert experience. According to Shannon sampling theorem, the maximum frequency fH of the signal can be obtained to ensure the effective reconstruction of industrial time series data X T The ideal sampling frequency f, where f≥2fH, is set as the optimal sampling frequency f=2fH for industrial time series data in this embodiment, which is a constant parameter.

[0066] Step S2, combining a data interpolation method and a data preprocessing method to perform data enhancement on the industrial time series data.

[0067] Specifically, the data interpolation method is used to convert the industrial time series data X with a length of N into T Expand to length N * Intermediate data Wherein, the data interpolation method is linear interpolation, cubic interpolation or Lagrange interpolation;

[0068]

[0069] In the formula, Indicates the data interpolation method;

[0070] In order to further improve the diversity of data, the image preprocessing methods in computer vision (such as rotation, jitter, scaling, reordering, etc.) are introduced into the time series data processing. Data preprocessing is performed to obtain industrial time series data after data enhancement; wherein the data preprocessing includes data rotation, data jitter, data scaling and data rearrangement.

[0071] Assume that the data before time series data preprocessing is The preprocessed data is X' T ;

[0072] Intermediate data generated by data interpolation Four data operations are performed: data rotation, data dithering, data scaling, and data rearrangement, among which:

[0073] Data rotation operation:

[0074]

[0075] Where R is the rotation parameter set.

[0076] Data jitter operation:

[0077]

[0078] Where, ∈ is the set of jitter parameters, ∈ trepresents the sum of the superimposed noise at each time t.

[0079] Data scaling operations:

[0080]

[0081] Where W is the amplitude scaling operation parameter, is a sequence, when they are a uniform value α, the data are scaled synchronously; or it can be obtained by cubic spline interpolation method, first through a normal distribution with a standard deviation of σ To select nodes u={u1,u2,…,u t}, and then use cubic spline interpolation CS(u) to find α t :

[0082]

[0083] Data rearrangement operation:

[0084]

[0085] Where i, j, k represent the first index slice of each window, so each slice is selected only once, using slices of uniform window size, w represents the slice window size, Where n is the total number of slices.

[0086] In the intermediate data In the process of performing four data operations, namely data rotation, data dithering, data scaling and data rearrangement, each operation has two possibilities: use / not use, for a total of 16 possibilities. For each data t∈[1,N * ], generates a random integer operator Ind(i) = random{0,1} of 0 or 1, corresponding to the usage of each data operation, through the matrix To indicate which types of data operations are taken in the preprocessing operation;

[0087] The data preprocessing operation is described as:

[0088]

[0089] In the formula, Represents data rotation operation, Indicates data jitter operation, represents a data scaling operation, Represents a data shuffle operation.

[0090] Combining data interpolation and data preprocessing, the new data obtained after the final time series data enhancement is:

[0091]

[0092] In the formula, F(·) WI represents data augmentation, represents the data preprocessing strategy, Represents the data interpolation strategy.

[0093] This embodiment combines the data interpolation method and the data preprocessing method to perform data enhancement on industrial time series data, thereby improving the richness and diversity of data samples, which is beneficial to parameter convergence and anti-overfitting in subsequent deep learning-based model training.

[0094] In step S3, a 1D single-layer convolutional neural network model with a temporal attention mechanism is used to perform anomaly detection on the industrial time series data after data enhancement.

[0095] Step S3 may specifically include steps S31 to S33, wherein:

[0096] Step S31, adopting the strategy of neighboring point label aggregation, by calculating the time domain aggregation feature index, the label information of the original data set is migrated to the enhanced data set.

[0097] Specifically, for the data element x in the original industrial time series dataset i , data element x i The time domain aggregation feature index during training is expressed as:

[0098]

[0099] Among them, N i is the neighborhood of node i in industrial time series data, x j is the neighborhood value, LeakyReLU(·) is a nonlinear activation function used in the training process, a ij is the aggregate score between data elements i and j;

[0100] Neighborhood normalization is performed through the softmax function to calculate the aggregation score of each neighborhood point;

[0101]

[0102] Among them, d(x i ,x j ) is the data element x i ,x j The distance between

[0103] Based on the aggregated scores, the labels in the original data are transferred to the enhanced data points.

[0104] Step S32, using the migrated label information in combination with the temporal attention mechanism, trains a 1D single-layer convolutional neural network model.

[0105] It is understandable that compared with the recurrent neural network model framework generally used for time series data in the fields of machine learning and deep learning, in the present invention, taking into account the requirements of industrial application fields for real-time response, a 1D single-layer convolutional neural network model with a time attention mechanism is adopted to achieve fast industrial time series data anomaly detection. Among them, the 1D single-layer convolutional neural network model (1DSingle-Layer ConvolutionalNeural NetworkModel) is a convolutional neural network structure specifically used to process one-dimensional data. Since only one layer of convolution is used, the computational complexity of the model is relatively low, which is suitable for processing large-scale time series data and meeting real-time requirements.

[0106] In the training process of the 1D single-layer convolutional neural network model, the enhanced data set that meets the neighborhood feature requirements is trained by aggregating feature indexes in the time domain. The data in the training data is automatically assigned labels that are consistent with the original data, and through this process, the time domain feature information is integrated into the training model, realizing the introduction of a temporal attention mechanism.

[0107] The loss function in the training process can be trained in various forms without special requirements. The present invention uses the root mean square error (RMSE) as the loss function for training.

[0108] Step S33, convert the time series data into two-dimensional data through the trained 1D single-layer convolutional neural network model, extract features through the convolution layer, and then output the distribution position of the abnormal data through maximum pooling and softmax classification.

[0109] In this embodiment, in order to ensure the detection speed and real-time performance of the anomaly detection model, the following Figure 2 A 1D single-layer convolutional neural network model is shown, which converts time series data into a horizontal axis with time T and a vertical axis with The two-dimensional data is directly input into a 1D single-layer convolutional neural network to divide the time series data into equal slices and extract its graphic features. The feature graphs are then concat-connected and sent to a maximum pooling layer for softmax classification to output the possible distribution locations and results of abnormal situations.

[0110] Reference Figure 2 , the structure of the 1D single-layer convolutional neural network model at least includes an input layer, a single-layer convolutional network (1D convolutional layer), a feature map concatenation, and an output layer. Among them:

[0111] Input layer: Convert the time series data into a two-dimensional form, where the horizontal axis represents the time step and the vertical axis represents the characteristic value of the data point (if the data is univariate, the vertical axis has only one feature; if it is multivariate, the vertical axis has multiple features).

[0112] 1D convolution layer: Use a one-dimensional convolution kernel to slide on the time dimension to extract local features of the time series. Since only one layer of convolution is used, the computational complexity of the model is low, which is conducive to real-time processing.

[0113] Feature map concatenation: The feature maps obtained after each subsequence passes through the convolutional layer are concatenated to form a complete feature representation.

[0114] Max pooling layer: The feature map is reduced in dimension through the max pooling layer, which reduces the number of parameters and extracts the most important features.

[0115] Output layer: Finally, the output of the pooling layer is fed into a softmax classifier to output the probability of each time step being an anomaly.

[0116] The present invention uses a 1D convolutional neural network to process time series data input, and further uses a softmax classification method to detect abnormal situations through feature extraction, converting time series problems into graphic features, improving detection efficiency, and being simple and convenient to use. In the model training process, the present invention introduces a strategy of time domain aggregation feature index, which not only preliminarily extracts the time series features of the data in the training data set, but also realizes automatic labeling of the enhanced data, reducing the labor cost and time cost of manual labeling, and improving the model training efficiency.

[0117] In a preferred embodiment of the present invention, after collecting industrial time series data in the industrial production process in step S1, data amplification can also be performed in the following manner:

[0118] For the discrete sparse time series data X that has been collected T ={x1, x2, ..., x i}, the original data length is N, and the discrete sampling time series index is T = {1, 2, ..., t}. In order to meet the requirements of the optimal sampling frequency, it is necessary to expand the data capacity. According to the requirements of the ideal sampling frequency f, the newly synthesized data through the data enhancement method is defined as Its data length is expanded to N * =2N*f-f+1, the expanded time series index becomes in:

[0119]

[0120] In this embodiment, after determining the optimal sampling frequency of industrial time series data, the originally collected discrete sparse time series data is amplified to a data capacity that meets the requirements of the optimal sampling frequency according to the optimal sampling frequency, providing a richer data basis for subsequent anomaly detection. This method not only improves the density and continuity of data, but also helps to improve the accuracy and reliability of anomaly detection.

[0121] Figure 3 The structural block diagram of the industrial time series data anomaly detection system based on data enhancement provided by the present invention is shown in FIG. Figure 3 ,The industrial time series data anomaly detection system based on data enhancement includes:

[0122] The data acquisition module 301 is used to collect industrial time series data in the industrial production process;

[0123] A data enhancement module 302, used to enhance the industrial time series data by combining a data interpolation method and a data preprocessing method;

[0124] The anomaly detection module 303 is used to perform anomaly detection on the industrial time series data after data enhancement by using a 1D single-layer convolutional neural network model with a temporal attention mechanism.

[0125] The industrial time series data anomaly detection system based on data enhancement provided by the present invention executes the industrial time series data anomaly detection method based on data enhancement provided by the above-mentioned embodiments through the above-mentioned modules. The industrial time series data anomaly detection method based on data enhancement has been described in detail in the above-mentioned embodiments, and this embodiment will not be repeated here.

[0126] Figure 4 The structural block diagram of the electronic device provided by the present invention is as follows: Figure 4 As shown, the present invention also provides an electronic device, and the electronic device 400 can be a computing device such as a mobile terminal, a desktop computer, a notebook, a palm computer, and a server. The electronic device 400 includes a processor 401 and a memory 402, wherein the memory 402 stores an industrial time series data anomaly detection program 404 based on data enhancement.

[0127] In some embodiments, the memory 402 may be an internal storage unit of a computer device, such as a hard disk or memory of the computer device. In other embodiments, the memory 402 may also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. Furthermore, the memory 402 may also include both an internal storage unit of the computer device and an external storage device. The memory 402 is used to store application software and various types of data installed on the computer device, such as program codes for installing the computer device. The memory 402 can also be used to temporarily store data that has been output or is to be output. In one embodiment, when the industrial time series data anomaly detection program 404 based on data enhancement is executed by the processor 401, the following steps are implemented:

[0128] Collect industrial time series data in industrial production processes;

[0129] Combining a data interpolation method and a data preprocessing method, the industrial time series data is enhanced;

[0130] A 1D single-layer convolutional neural network model with a temporal attention mechanism is used to perform anomaly detection on industrial time series data after data enhancement.

[0131] In some embodiments, the processor 401 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 402, such as executing an industrial time series data anomaly detection program based on data enhancement.

[0132] This embodiment further provides a computer-readable storage medium, on which is stored an industrial time series data anomaly detection program based on data enhancement. When the industrial time series data anomaly detection program based on data enhancement is executed by a processor, the following steps are implemented:

[0133] Collect industrial time series data in industrial production processes;

[0134] Combining a data interpolation method and a data preprocessing method, the industrial time series data is enhanced;

[0135] A 1D single-layer convolutional neural network model with a temporal attention mechanism is used to perform anomaly detection on industrial time series data after data enhancement.

[0136] In summary, the industrial time series data anomaly detection method and system based on data enhancement provided by the present invention have the following beneficial effects compared with the prior art:

[0137] 1) The present invention increases the data sample capacity and the data sampling frequency by interpolating and enhancing the original industrial time series data, which plays a key role in restoring the key information contained in the real data from the enhanced data.

[0138] 2) The present invention combines data interpolation methods and data preprocessing methods to perform data enhancement on industrial time series data, thereby improving the richness and diversity of data samples, which is beneficial to parameter convergence and anti-overfitting in subsequent deep learning-based model training.

[0139] 3) In the process of model training, the present invention introduces the strategy of time-domain aggregation feature index, which not only preliminarily extracts the temporal features of the data in the training data set, but also realizes the automatic labeling of the enhanced data, reduces the labor cost and time cost of manual labeling, and improves the efficiency of model training;

[0140] 4) The present invention uses a 1D convolutional neural network to process time series data input, and further uses a softmax classification method to detect abnormal situations through feature extraction, converting time series problems into graphic features, thereby improving detection efficiency and being simple and convenient to use.

[0141] The above-described embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting anomalies in industrial time series data based on data enhancement, characterized in that: include: Collect industrial time series data in industrial production processes; Combining a data interpolation method and a data preprocessing method, the industrial time series data is enhanced; A 1D single-layer convolutional neural network model with a temporal attention mechanism is used to perform anomaly detection on industrial time series data after data enhancement.

2. The method for detecting anomalies in industrial time series data based on data enhancement according to claim 1, characterized in that: The process of collecting industrial time series data in the industrial production process also includes: Based on Shannon sampling theorem and prior knowledge, the optimal sampling frequency of industrial time series data is determined; wherein the industrial time series data includes voltage, current, temperature, humidity and speed.

3. The method for detecting anomalies in industrial time series data based on data enhancement according to claim 1, characterized in that: The method of combining the data interpolation method and the data preprocessing method to perform data enhancement on the industrial time series data includes: Using the data interpolation method, the industrial time series data X with a length of N is T Expand to length N * Intermediate data Wherein, the data interpolation method is linear interpolation, cubic interpolation or Lagrange interpolation; In the formula, Indicates the data interpolation method; Intermediate data generated by interpolation Data preprocessing is performed to obtain industrial time series data after data enhancement; wherein the data preprocessing includes data rotation, data jitter, data scaling and data rearrangement.

4. The method for detecting anomalies in industrial time series data based on data enhancement according to claim 3 is characterized in that: The intermediate data generated by interpolation Perform data preprocessing to obtain industrial time series data after data enhancement, including: Assume that the data before time series data preprocessing is The preprocessed data is X' T ; Intermediate data generated by data interpolation Four data operations are performed: data rotation, data jitter, data scaling, and data rearrangement. Each operation has two possibilities: use or not use, for a total of 16 possibilities. For each data t∈[1,N * ], generates a random integer operator Ind(i) = random{0,1} of 0 or 1, corresponding to the usage of each data operation, through the matrix To indicate which types of data operations are taken in the preprocessing operation; The data preprocessing operation is described as: In the formula, Represents data rotation operation, Indicates data jitter operation, represents a data scaling operation, Represents a data shuffle operation.

5. The method for detecting anomalies in industrial time series data based on data enhancement according to claim 1, characterized in that: The 1D single-layer convolutional neural network model with a temporal attention mechanism is used to perform anomaly detection on industrial time series data after data enhancement, including: The strategy of aggregating labels of adjacent points is adopted to migrate the label information of the original dataset to the enhanced dataset by calculating the time domain aggregation feature index. Using the migrated label information and the temporal attention mechanism, a 1D single-layer convolutional neural network model is trained. The time series data is converted into two-dimensional data through the trained 1D single-layer convolutional neural network model. The features are extracted through the convolution layer, and then the distribution position of the abnormal data is output after maximum pooling and softmax classification.

6. The method for detecting anomalies in industrial time series data based on data enhancement according to claim 5 is characterized in that: The strategy of aggregating labels of adjacent points is adopted to migrate the label information of the original data set to the enhanced data set by calculating the time domain aggregation feature index, including: For the data element x in the original industrial time series dataset i , data element x i The time domain aggregation feature index during training is expressed as: Among them, N i is the neighborhood of node i in industrial time series data, x j is the neighborhood value, LeakyReLU(·) is a nonlinear activation function used in the training process, a ij is the aggregate score between data elements i and j; Neighborhood normalization is performed through the softmax function to calculate the aggregation score of each neighborhood point; Among them, d(x i ,x j ) is the data element x i ,x j The distance between Based on the aggregated scores, the labels in the original data are transferred to the enhanced data points.

7. An industrial time series data anomaly detection system based on data enhancement, characterized in that: include: Data acquisition module, used to collect industrial time series data in the industrial production process; A data enhancement module, used to enhance the industrial time series data by combining a data interpolation method and a data preprocessing method; The anomaly detection module is used to detect anomalies in industrial time series data after data enhancement by using a 1D single-layer convolutional neural network model with a temporal attention mechanism.

8. An electronic device, It is characterized in that comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the industrial time series data anomaly detection method based on data enhancement as described in any one of claims 1 to 6 above.

9. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the industrial time series data anomaly detection method based on data enhancement as described in any one of claims 1 to 6 above.

Images