A multi-scale instrument fault diagnosis method based on ensemble learning

Abstract

This invention discloses a multi-scale instrument fault diagnosis method based on ensemble learning, belonging to the field of fault diagnosis technology, and includes the following steps: Step S1, multi-scale fault information conversion; Step S2, obtaining instrument fault diagnosis results from multiple models; Step S3, fusing multiple diagnostic decisions from fault information at different scales. This invention employs the aforementioned multi-scale instrument fault diagnosis method based on ensemble learning, integrating different diagnostic methods to improve diagnostic accuracy. It addresses the problems of cumbersome preprocessing and low diagnostic rates in ensemble learning-based diagnostic methods, which often involve complex multi-scale fault sequences containing diverse fault information. This achieves convenient, practical, and accurate instrument fault diagnosis, thus improving the reliability of instrument applications in the industry.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and in particular to a multi-scale instrument fault diagnosis method based on ensemble learning. Background Technology

[0002] Instrumentation is a crucial component of the information technology industry, playing a key role in the acquisition, storage, transmission, processing, and utilization of information. Instrument fault diagnosis not only affects production efficiency but also directly impacts equipment and personnel safety, preventing or eliminating potential catastrophic accidents. Timely instrument fault diagnosis is critical to ensuring the normal operation of automated systems, effectively reducing downtime and maintenance costs. Current methods focus on integrating different diagnostic methods to improve performance; however, the complex preprocessing steps of different methods make ensemble learning-based diagnostic methods cumbersome. Furthermore, current methods do not consider multi-scale integration to improve instrument fault diagnosis performance, resulting in insufficient diagnostic capabilities. Therefore, it is necessary to consider both multi-scale and ensemble learning for instrument fault diagnosis. Summary of the Invention

[0003] The purpose of this invention is to provide a multi-scale instrument fault diagnosis method based on ensemble learning, which solves the problems of cumbersome preprocessing and low diagnosis rate in ensemble methods for instrument fault diagnosis.

[0004] To achieve the above objectives, this invention provides a multi-scale instrument fault diagnosis method based on ensemble learning, comprising the following steps:

[0005] Step S1: The fault sequence is compressed into multi-granularity subsequences using an average pooling layer, so that multi-scale fault information can be extracted from the subsequences and the original sequence with a fixed kernel size, thereby realizing the multi-scale fault information conversion.

[0006] Step S2: Input the time series of multi-scale fault information into multiple identical fully convolutional diagnostic models to obtain the instrument fault diagnosis results of multiple models;

[0007] Step S3: Integrate multiple diagnostic decisions from fault information at different scales, and recalculate the final instrument fault mode probability distribution through a fully connected layer.

[0008] Preferably, average pooling is used in step S1. k-1 Extracting multi-scale fault information using average pooling. k-1 As shown in the following formula:

[0009]

[0010] Among them, F kjThis indicates that the average pooling operation is used. k-1 The j-th value of the obtained vector, the length of the subsampling pooling domain (1×k matrix), and the step size of the average pooling process are set to k, x s This represents the s-th value of the input sample x. H represents the length of the input sample time window.

[0011] Preferably, in step S1, n average pooling layers are used for multi-scale fault information conversion, and n is obtained by the following formula:

[0012]

[0013] Where l represents the length of the time series sample, m is the maximum value of n, set to 5, and f represents the size of the first filter of the original model.

[0014] Preferably, step S2 includes the following specific steps:

[0015] Multiple time series data transformed by scale fault information are input into n fully convolutional network diagnostic models. Each fully convolutional network diagnostic model consists of three feature extractors, one global average pooling layer, and one softmax layer.

[0016] The feature extractor for FCN is shown in the following equation:

[0017]

[0018] in, BN and ReLU represent the convolution operator, batch normalization layer, and corrected linear unit, respectively. W con b represents the kernel parameter of the fully convolutional layer. con This represents the bias of the convolution kernel; the filter sizes of the three convolution operators are 128, 256, and 128, respectively, and the one-dimensional kernel sizes of the three convolution operators are 8, 5, and 3, respectively;

[0019] Each diagnostic model obtains a diagnostic result through a softmax layer.

[0020] Preferably, step S3 includes the following specific steps:

[0021] The n diagnostic decisions from fault information at different scales are fused together to form the final fault mode probability distribution.

[0022] Given the i-th diagnostic decision is Where c represents the number of fault modes, the fusion output of n diagnostic decisions is shown in the following formula:

[0023]

[0024] The probability of the z-th failure mode is obtained by the following method:

[0025] v = WF + b;

[0026]

[0027] in, These are the network parameters of the fully connected layer, where b represents the offset, c represents the number of fault modes, e represents the natural constant, v represents the fault probability vector, and d represents the number of cumulative calculations.

[0028] Therefore, the present invention adopts the above-mentioned multi-scale instrument fault diagnosis method based on ensemble learning, which integrates different diagnostic methods to improve diagnostic accuracy. This solves the problems of cumbersome preprocessing and low diagnostic rate of integrated instrument fault diagnosis methods, which are based on ensemble learning. It achieves convenient, practical and accurate instrument fault diagnosis, which is conducive to improving the reliability of instrument industrial applications.

[0029] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating an embodiment of a multi-scale instrument fault diagnosis method based on ensemble learning according to the present invention.

[0031] Figure 2 This is a schematic diagram of a fully convolutional network diagnostic model, representing an embodiment of a multi-scale instrument fault diagnosis method based on ensemble learning according to the present invention.

[0032] Figure 3 This is a probability density distribution diagram of the accuracy gain of an embodiment of the multi-scale instrument fault diagnosis method based on ensemble learning of the present invention. Detailed Implementation

[0033] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0034] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0035] Example

[0036] Please see Figure 1-3 This invention provides a multi-scale instrument fault diagnosis method based on ensemble learning, comprising the following steps:

[0037] Step S1: The fault sequence is compressed into multi-granularity subsequences using an average pooling layer, so that multi-scale fault information can be extracted from the subsequences and the original sequence with a fixed kernel size, thereby realizing the multi-scale fault information conversion.

[0038] The goal of this method is to maintain the usability of the multi-scale integrated instrument diagnostic model while extracting multi-scale fault information. It employs an operation similar to spatial pyramid pooling layers, and uses average pooling to preserve local fault information from the time series as much as possible. k-1 Instead of max pooling, multi-scale fault information is extracted. Furthermore, the transformed fault information is sent separately to multiple original models, rather than concatenating them into a one-dimensional vector. Average pooling is used instead. k-1 As shown in the following formula:

[0039]

[0040] Among them, F kj This indicates that the average pooling operation is used. k-1 The j-th value of the obtained vector, the length of the subsampling pooling domain (1×k matrix), and the step size of the average pooling process are set to k, x s This represents the s-th value of the input sample x. H represents the length of the input sample time window.

[0041] In the multi-scale fault information transformation stage, n average pooling layers are used for multi-scale fault information transformation. n is the only hyperparameter that needs to be set in this method, and a simple but effective rule is set for it. A maximum value is set for n, and attempts are made to set n to a larger value within the constraints of the integrated model. n is obtained through the following formula:

[0042]

[0043] Where l represents the length of the time series sample, m is the maximum value of n, set to 5, and f represents the size of the first filter of the original model.

[0044] After the multi-scale fault information transformation stage, the original fault sequence is transformed into a fault sequence of n scales. This enables the original model to extract multi-scale fault information, which can help extract multi-scale fault feature information without changing the structure and hyperparameters of the integrated model.

[0045] Step S2: Input the time series data after multi-scale fault information conversion into multiple identical fully convolutional diagnostic models to obtain the instrument fault diagnosis results from multiple models. Specific steps include:

[0046] Multiple time series data, after being transformed by scale-based fault information, are input into n fully convolutional network (FCN) diagnostic models. Each FCN diagnostic model consists of three feature extractors, a global average pooling layer, and a softmax layer. The feature extractor of the FCN is shown in the following equation:

[0047]

[0048] in, BN and ReLU represent the convolution operator, batch normalization layer, and corrected linear unit, respectively. W con b represents the kernel parameter of the fully convolutional layer. con This represents the bias of the convolution kernel. The filter sizes for the three convolution operators are 128, 256, and 128, respectively, and the one-dimensional kernel sizes for the three convolution operators are 8, 5, and 3, respectively.

[0049] Each diagnostic model, such as Figure 2 As shown, each diagnostic model obtains a diagnostic result through a softmax layer.

[0050] Step S3: Integrate multiple diagnostic decisions from fault information at different scales, and recalculate the final instrument fault mode probability distribution through a fully connected layer. Specific steps include:

[0051] The n diagnostic decisions from fault information at different scales are fused together to form the final fault mode probability distribution.

[0052] Given the i-th diagnostic decision is Where c represents the number of failure modes. The fusion output of n diagnostic decisions is shown in the following formula:

[0053]

[0054] The probability of the z-th failure mode is obtained by the following method:

[0055] v = WF + b;

[0056]

[0057] in, These are the network parameters of the fully connected layer, where b represents the offset, c represents the number of fault modes, e represents the natural constant, v represents the fault probability vector, and d represents the number of cumulative calculations.

[0058] Figure 2 and Figure 3 The probability density distribution of the fully convolutional network diagnostic model and accuracy gain provided in this embodiment is shown. Comparing this method with a fully convolutional network diagnostic model that does not consider multi-scale integration demonstrates the applicability and effectiveness of this method for classification results, as well as its higher diagnostic accuracy.

[0059] Therefore, the present invention adopts the above-mentioned multi-scale instrument fault diagnosis method based on ensemble learning, which integrates different diagnostic methods to improve diagnostic accuracy. This solves the problems of cumbersome preprocessing and low diagnostic rate of integrated instrument fault diagnosis methods, which are based on ensemble learning. It achieves convenient, practical and accurate instrument fault diagnosis, which is conducive to improving the reliability of instrument industrial applications.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-scale instrument fault diagnosis method based on ensemble learning, characterized in that, The method comprises the following steps: Step S1, adopting an average pooling layer to compress the fault sequence into multi-granularity sub-sequences, so as to extract multi-scale fault information from the sub-sequences and the original sequence to realize multi-scale fault information conversion; Step S2, inputting the time sequence converted by the multi-scale fault information into a plurality of same full convolution diagnosis models to obtain instrument fault diagnosis results of the plurality of models; Step S3, fusing a plurality of diagnosis decisions from different scale fault information, and obtaining a final instrument fault mode probability distribution through re-computation by a full connection layer; In step S1, n average pooling layers are used for multi-scale fault information conversion, and n is obtained by the following formula: Wherein, l represents the length of the time sequence sample, m is the maximum value of n and is set to 5, and f represents the size of the first filter of the original model; The specific steps of step S2 include: The plurality of time sequences converted by the scale fault information are input into n full convolution network diagnosis models, each of which is composed of three feature extractors, a global average pooling layer and a softmax layer; The feature extractor of the FCN is shown in the following formula: wherein, BN and ReLU represent a convolution operator, a batch normalization layer and a rectified linear unit, respectively, W con represents a convolution kernel parameter of a full convolution layer, b con represents a bias of the convolution kernel; the filter sizes of the three convolution operators are 128, 256 and 128, respectively, and the one-dimensional kernel sizes of the three convolution operators are 8, 5 and 3, respectively; Each diagnosis model obtains a diagnosis result through the softmax layer.

2. The multi-scale instrument fault diagnosis method based on ensemble learning according to claim 1, characterized in that, The average pooling operation Average pooling is used in step S1 k-1 Extracting multi-scale fault information, the average pooling operation Average pooling k-1 As shown in the following formula: where F kj denotes the jth value of the vector obtained by the average pooling operation Average pooling k-1 denotes the jth value of the vector obtained by the average pooling operation Average pooling s denotes the sth value of the input sample x, the input H denotes the length of the input sample time window.

3. The multi-scale instrument fault diagnosis method based on ensemble learning according to claim 2, characterized in that, The specific steps of step S3 include: The n diagnosis decisions from different scale fault information are fused together to form a final fault mode probability distribution; Given the ith diagnosis decision as where c denotes the number of fault modes, the fusion output of n diagnosis decisions is given by The probability of the zth fault mode is obtained by the following formula: v = WF + b; wherein, are network parameters of the fully connected layer, b represents an offset, c represents the number of failure modes; e represents a natural constant; v represents a failure probability vector; d represents the number of accumulative calculations.

Images