CN-121997149-A - Valve fault diagnosis method and system based on artificial intelligence

CN121997149ACN 121997149 ACN121997149 ACN 121997149ACN-121997149-A

Abstract

The invention discloses a valve fault diagnosis method and system based on artificial intelligence, and relates to the technical field of valves. The method comprises the steps of constructing a mechanism semantic skeleton of a valve, obtaining a valve theoretical response sequence based on the mechanism semantic skeleton, collecting a historical valve fault response sequence, calculating the difference between the historical valve fault response sequence and the valve theoretical response sequence through an improved dynamic time warping method, calculating to obtain a valve fault residual sequence, extracting characteristics of the valve fault residual sequence to obtain valve fault characteristic data, training a time sequence convolution network model based on an attention mechanism by taking the valve fault characteristic data as a training set to obtain a trained time sequence convolution network model, collecting real-time valve fault characteristic data, inputting the real-time valve fault characteristic data into the trained time sequence convolution network model, outputting to obtain a valve fault result, calculating to obtain final fault confidence and classifying fault grades based on the valve fault result.

Inventors

ZHENG ZHONGXING
QI HUAN
CHEN SHUANGSHUANG
SHEN WEIGUANG
LIU YU
ZHOU YAQUN

Assignees

浙江新欧自控仪表有限公司

Dates

Publication Date: 20260508
Application Date: 20260410

Claims (10)

1. The valve fault diagnosis method based on artificial intelligence is characterized by comprising the following steps of: Step S1, constructing a mechanism semantic skeleton of a valve, and obtaining a valve theoretical response sequence based on the mechanism semantic skeleton; Step S2, collecting a historical valve fault response sequence, calculating the difference between the historical valve fault response sequence and a valve theoretical response sequence through an improved dynamic time warping method, and calculating to obtain a valve fault residual sequence; Step S3, training the time sequence convolution network model based on the attention mechanism by taking the valve fault characteristic data as a training set to obtain a trained time sequence convolution network model; And S4, acquiring real-time valve fault characteristic data, inputting the real-time valve fault characteristic data into a trained time sequence convolution network model, outputting and obtaining a valve fault result, calculating and obtaining final fault confidence and dividing fault grades based on the valve fault result, and therefore diagnosing the valve fault.
2. The valve fault diagnosis method based on artificial intelligence according to claim 1, wherein the construction of the mechanism semantic skeleton of the valve comprises the following specific steps: The mechanism framework comprises valve core valve plate motion constraint, valve rod force transmission chain, actuating mechanism output characteristic, sealing compression section characteristic and flow passage resistance characteristic, and the valve is disassembled into 3 action stages, namely starting to overcome static friction, continuously changing displacement and stopping sealing compression formation.
3. The method for diagnosing valve faults based on artificial intelligence according to claim 2, wherein the calculating of the difference between the historical valve fault response sequence and the theoretical valve response sequence through the improved dynamic time warping method is carried out, and the valve fault residual sequence is obtained through calculation, and the method comprises the following steps: Calculating theoretical response sequence of valve in ith action stage With historical valve failure response sequences Is the initial regular distance of (2) I is an action phase index; Performing phase compensation on the valve theoretical response sequence to obtain a valve theoretical response sequence after phase compensation; Calculating a theoretical response sequence of the valve after phase compensation With historical valve failure response sequences And constructing a distance matrix; Resampling based on the distance matrix to obtain an aligned historical valve fault response sequence (T) sequence of theoretical responses of aligned valve ; Aligned historical valve fault response sequences (T) sequence of theoretical responses of aligned valve Subtracting and normalizing to obtain a valve fault residual sequence 。
4. The method for diagnosing valve faults based on artificial intelligence as claimed in claim 3, wherein the phase compensation of the valve theoretical response sequence comprises the following steps: Theoretical response sequence of valve in ith action stage From 0 to 2 pi by phase angle theta Step length offset, obtain valve theoretical response sequence after offset Calculating the regular distance after each offset Screening out the products meeting the requirements < And selecting the regular distance after offset Is the minimum distance in (2) Corresponding phase angle As the optimal phase compensation angle, calculating an optimal time offset based on the optimal phase compensation angle Performing phase compensation on the valve theoretical response sequence to obtain a valve theoretical response sequence after phase compensation , = 。
5. The method for diagnosing valve faults based on artificial intelligence as claimed in claim 4, wherein the distance matrix is constructed specifically as follows: constructing a distance matrix, wherein the elements of the distance matrix are as follows: ; Wherein, the Representing the behavior j, the element value listed as j2, For the derivative of the jth 2 sampling points of the ith action phase in the historical valve failure response sequence, The method is characterized in that the method comprises the steps of taking the derivative of a jth sampling point in an ith action stage in a valve theoretical response sequence, J is the index of the sampling point of the valve theoretical response sequence, J is the length of the valve theoretical response sequence, J2 is the index of a historical valve fault response sequence, and J2 is the length of the historical valve fault response sequence.
6. The method for diagnosing valve faults based on artificial intelligence as claimed in claim 5, wherein the step of extracting the characteristics of the valve fault residual sequence to obtain valve fault characteristic data comprises the following steps: Valve failure residual sequence for the ith action phase Extracting global time domain features and local sliding window features, and orderly splicing the global time domain features and the local sliding window features according to 3 action phases to form complete residual features; And calculating a residual structure template and a residual evolution track template based on residual characteristics, and finally obtaining valve fault characteristic data.
7. The method for diagnosing valve faults based on artificial intelligence of claim 6, wherein the training of the time sequence convolution network model based on an attention mechanism by taking valve fault characteristic data as a training set is carried out to obtain a trained time sequence convolution network model, and the method comprises the following specific steps of: Constructing a time sequence convolution network model based on an attention mechanism; the total loss function of the time sequence convolution network model is as follows: ; wherein, loss is the total Loss function value, The loss function is classified for the failure mechanism, The loss function is classified for the structural part, The task weight coefficient is classified; The time sequence convolution network model is trained by using an Adam optimizer, valve fault characteristic data is used as a training set to train the time sequence convolution network model based on an attention mechanism, and a trained time sequence convolution network model is obtained.
8. The method for diagnosing valve faults based on artificial intelligence as claimed in claim 7, wherein the time sequence convolution network model based on an attention mechanism is constructed specifically as follows: The time sequence convolution network model based on the attention mechanism comprises an input layer, a time sequence convolution module, an ECA attention module, a pyramid pooling layer, a full connection layer and an output layer.
9. The method for diagnosing valve faults based on artificial intelligence according to claim 8 is characterized in that the final fault confidence is calculated based on a valve fault result, specifically: final fault confidence Is calculated according to the formula: ; Wherein, the In order to achieve a final degree of failure confidence, As a result of the core fault confidence level, Is a fault-free correction term.
10. An artificial intelligence based valve fault diagnosis system comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to carry out the steps of the method according to any one of claims 1-9.

Description

Valve fault diagnosis method and system based on artificial intelligence Technical Field The invention relates to the technical field of valves, in particular to a valve fault diagnosis method and system based on artificial intelligence. Background The valve is used as a core key component for fluid transportation, pressure regulation and medium on-off control in an industrial production process, and is widely applied to the fields of petrochemical industry, electric power, metallurgy, municipal water supply and the like, and the stability of the running state of the valve is directly related to the continuous operation of a production system, the precise control of technological parameters and the safety protection of on-site production. In the long-term service process of the valve, the valve is influenced by multiple factors such as medium corrosion, mechanical abrasion, working condition fluctuation and the like, various faults such as valve rod eccentric wear, sealing surface damage, valve core blocking and the like are easy to occur, if the valve cannot be identified and treated in time, the production efficiency is reduced, the energy consumption is increased, and serious economic losses are caused by medium leakage, equipment shutdown and even safety accidents, so that the accurate diagnosis research of the valve faults is carried out, the early identification and state evaluation of the faults are realized, and the valve is an important requirement in the field of industrial equipment state monitoring. The traditional valve fault diagnosis technology is mainly based on vibration amplitude variation for analysis, vibration signals generated in the valve operation process are continuously collected by arranging vibration sensors at key structure parts such as a valve body and a valve rod, vibration amplitude in the vibration signals is extracted, the vibration amplitude is compared with a preset amplitude threshold, and when the vibration amplitude exceeds the threshold range, the valve is judged to have mechanical abnormal faults, so that the diagnosis and judgment of the valve faults are completed. However, the traditional method only carries out fault discrimination by means of vibration amplitude, and cannot adapt to signal dynamic change characteristics under complex working conditions such as cooperative vibration of industrial field peripheral equipment, interference of environmental noise and the like, so that normal vibration amplitude fluctuation of a valve caused by complex working conditions cannot be effectively distinguished, abnormal vibration amplitude change caused by mechanical abnormality of the valve is avoided, finally, misjudgment rate of fault diagnosis is obviously increased, and stable diagnosis accuracy is difficult to maintain under complex working conditions. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a valve fault diagnosis method and system based on artificial intelligence, which solve the problems existing in the background art. In order to achieve the purpose, the invention is realized by the following technical scheme that the valve fault diagnosis method and system based on artificial intelligence comprises the following steps: Step S1, constructing a mechanism semantic skeleton of a valve, and obtaining a valve theoretical response sequence based on the mechanism semantic skeleton; Step S2, collecting a historical valve fault response sequence, calculating the difference between the historical valve fault response sequence and a valve theoretical response sequence through an improved dynamic time warping method, and calculating to obtain a valve fault residual sequence; Step S3, training the time sequence convolution network model based on the attention mechanism by taking the valve fault characteristic data as a training set to obtain a trained time sequence convolution network model; And S4, acquiring real-time valve fault characteristic data, inputting the real-time valve fault characteristic data into a trained time sequence convolution network model, outputting and obtaining a valve fault result, calculating and obtaining final fault confidence and dividing fault grades based on the valve fault result, and therefore diagnosing the valve fault. Preferably, the construction of the mechanism semantic skeleton of the valve comprises the following specific steps: The mechanism framework comprises valve core valve plate motion constraint, valve rod force transmission chain, actuating mechanism output characteristic, sealing compression section characteristic and flow passage resistance characteristic, and the valve is disassembled into 3 action stages, namely starting to overcome static friction, continuously changing displacement and stopping sealing compression formation. Preferably, the calculating the difference between the historical valve fault response sequence and the theoretical valve response sequence by the impro