CN-121997231-A - Flow meter predictive maintenance method and system based on artificial intelligence

CN121997231ACN 121997231 ACN121997231 ACN 121997231ACN-121997231-A

Abstract

The application relates to the technical field of artificial intelligence and discloses a flow meter predictive maintenance method and system based on artificial intelligence, wherein the method firstly learns multidimensional operation parameters through an artificial intelligence model to preliminarily predict electrode fault types; then, introducing a physical verification link based on an equivalent capacitance, comparing an AI prediction result with an ideal capacitance value determined by a fault type, thereby verifying and correcting the preliminary diagnosis, and finally outputting a precise target fault type and confidence level; and finally, generating a personalized maintenance strategy according to the residual maintenance time of the fault and the target fault type. Therefore, the application can predict and identify different fault types of the electrode in advance and provide a targeted maintenance strategy.

Inventors

DU LIANG
SU XUEJIAN
XU SHUFAN
CHEN YANG
PU CHENG
ZHANG YAHUI

Assignees

天津迅尔科技股份有限公司
迅尔仪表(沧州)有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. An artificial intelligence-based flow meter predictive maintenance method applied to a flow meter maintenance system, the method comprising: Acquiring a multi-dimensional parameter of the operation of the flowmeter, wherein the multi-dimensional parameter at least comprises one or more of a flow parameter, a conductivity parameter, a vibration parameter, an environment parameter, an instrument power supply parameter and a medium characteristic parameter; extracting characteristics of the collected multidimensional parameters to obtain static characteristics, dynamic characteristics and coupling characteristics; Inputting the extracted features into a pre-trained fault prediction model, and outputting an electrode fault prediction type and a corresponding confidence coefficient, wherein the electrode fault prediction type comprises an electrode water film covering fault, an electrode scaling covering fault, an electrode vibration loosening fault and an electrode corrosion damage fault; Performing capacitance verification on the electrode fault prediction type according to the electrode ideal capacitance value and the actual equivalent capacitance value corresponding to the electrode fault prediction type to obtain an electrode target fault type and a target confidence coefficient; According to the target fault type and the corresponding target confidence coefficient, a preset fault degradation model is called to simulate an evolution path of the fault under the current actual working condition, and the residual maintenance time of the fault is calculated; and matching and generating a targeted maintenance strategy from a preset strategy library according to the residual maintenance time of the fault and the target fault type.
2. The method of flow meter predictive maintenance of claim 1, wherein the static characteristic is a degree to which a parameter deviates from a preset normal range, including a degree of deviation of the parameter, the degree of deviation calculated as: ; Wherein, delta is the parameter deviation degree, x is the parameter real-time acquisition value, x 0 is the parameter preset standard value, x_max is the parameter allowable maximum value, x_min is the parameter allowable minimum value, and when delta is more than or equal to 20%, the characteristic of high deviation is marked.
3. The method of predictive maintenance of a flow meter of claim 1, wherein the dynamic characteristic is a time series trend of the parameter, including a rate of change of the parameter, the rate of change being obtained by calculating a slope of a change of the parameter within adjacent sliding time windows by a linear fitting algorithm, wherein a length of the sliding time windows is adaptively adjusted according to a period of the change of the parameter.
4. The method of maintaining flow meter predictability according to claim 1, wherein the coupling feature is a synchronous change association between multiple parameters, including correlation coefficients of parameter pairs, the inputting the extracted feature into a pre-trained fault prediction model, outputting an electrode fault prediction type and a corresponding confidence level, comprising: Invoking static features and dynamic features through a feature screening algorithm to perform initial prediction of fault types to obtain candidate electrode fault prediction types; matching the corresponding coupling characteristics according to the candidate electrode fault prediction type to obtain target coupling characteristics, wherein the target coupling characteristics comprise correlation coefficients of target parameter pairs; And respectively carrying out weight distribution and key information focusing on the static characteristics, the dynamic characteristics and the target coupling characteristics by utilizing a parallel attention mechanism network layer, and then outputting the fault prediction type and the corresponding confidence of the flowmeter electrode through model reasoning.
5. The method for maintaining the predictability of a flow meter according to claim 4, wherein when there are a plurality of candidate electrode fault prediction types, the weighting distribution and the key information focusing are performed on the static feature, the dynamic feature and the target coupling feature by using the parallel attention mechanism network layer, respectively, and then the fault prediction types and the corresponding confidence degrees of the flow meter electrodes are output through model reasoning, comprising: Extracting a correlation coefficient r of a coupling characteristic corresponding to each candidate fault prediction type, comparing the correlation coefficient r with a preset correlation coefficient threshold r 0 of the coupling characteristic of the fault prediction type, and reserving a target coupling characteristic with r being more than or equal to r 0 and the corresponding candidate fault prediction type to form an effective candidate set; Inputting the corresponding static characteristics, dynamic characteristics and target coupling characteristics of each candidate fault prediction type in the effective candidate set into a parallel attention mechanism network layer, carrying out weight distribution on the three types of characteristics through the network layer, and outputting the corresponding preliminary confidence coefficient of each candidate type; And selecting the candidate fault prediction type with the highest preliminary confidence coefficient and no lower than a preset value as a final fault prediction type, and if the preliminary confidence coefficient of all candidate types in the effective candidate set is lower than the preset value, outputting the candidate types with the first two confidence coefficients and the correlation coefficient details of the target coupling characteristics of the candidate types, triggering a manual rechecking process, and manually confirming the final fault type.
6. The method of maintaining predictability of a flow meter of claim 1, wherein performing a capacitive verification on the electrode fault prediction type based on an ideal electrode capacitance value and an actual equivalent capacitance value corresponding to the electrode fault prediction type to obtain an electrode target fault type and a target confidence level comprises: calculating the relative deviation percentage of the actual equivalent capacitance value and the ideal capacitance value; If the relative deviation percentage is determined to be smaller than a first threshold value, judging that the capacitance check is passed, and directly determining the electrode fault prediction type as the electrode target fault type; determining that the relative deviation percentage is larger than or equal to a first threshold value and smaller than a second threshold value, judging that the working condition is interfered, maintaining the original prediction type as the target fault type, and reducing the corresponding confidence coefficient; And if the relative deviation percentage is determined to be greater than or equal to a second threshold value, judging that the capacitance check is not passed, reducing the corresponding confidence coefficient to zero, and triggering a manual maintenance mechanism.
7. The method for predictive maintenance of a flow meter according to claim 1, wherein the step of calling a preset failure degradation model according to the target failure type and the corresponding target confidence level to simulate an evolution path of a failure under a current actual working condition, and calculating remaining maintenance time of the failure comprises: According to the target fault type, calling a preset fault degradation model to simulate an evolution path of the fault under the current actual working condition to obtain a residual maintenance time initial evaluation value; And introducing a confidence coefficient correction factor to correct the initial evaluation value of the remaining maintenance time to obtain the remaining maintenance time of the fault, wherein a corrected remaining maintenance time calculation formula is as follows: ; Wherein T is the residual maintenance time after correction, T 0 is the residual maintenance time of initial calculation of the model, alpha is a confidence coefficient correction coefficient, the value range is 0.3-0.8, and gamma is the target confidence coefficient.
8. The method of predictive maintenance of a flow meter of claim 7, wherein invoking a predetermined fault degradation model to simulate an evolving path of a fault under current actual conditions to obtain an initial evaluation of remaining maintenance time based on a target fault type comprises: the fault deterioration model is integrated with a real-time working condition dynamic adjustment factor when simulating an evolution path, wherein the dynamic adjustment factor comprises a medium temperature fluctuation coefficient, a pressure change coefficient and an environmental humidity influence coefficient, and a target evolution curve is obtained by calculating and correcting the evolution curve through weighting; And determining a residual maintenance time initial evaluation value according to the target evolution curve.
9. The method of predictive maintenance of a flow meter of claim 8, wherein modifying the evolution curve by a weighted calculation to obtain the target evolution curve comprises: Respectively distributing preset weights for a medium temperature fluctuation coefficient k 1 , a pressure change coefficient k 2 and an environmental humidity influence coefficient k 3 ; calculating the comprehensive adjustment coefficient K, wherein the calculation formula is Wherein Respectively is Corresponding preset weights; And correcting the time axis parameter t of the initial fault evolution curve according to the comprehensive adjustment coefficient K, wherein the correction formula is t '=tx (1+K), and if K is a negative value, t' =tx (1- |K|) to obtain a corrected target evolution curve.
10. A flow meter maintenance system comprising a memory for storing program code and a processor for invoking the program code to perform the method of any of claims 1 to 9.

Description

Flow meter predictive maintenance method and system based on artificial intelligence Technical Field The invention relates to the technical field of artificial intelligence, in particular to a flow meter predictive maintenance method and system based on artificial intelligence. Background In industrial production and various fluid monitoring scenes, the flowmeter is a vital device, and stable operation and accurate measurement of the flowmeter play a key role in optimizing production flow, reasonably utilizing resources and guaranteeing product quality. The traditional flowmeter maintenance method is mainly periodic maintenance and passive maintenance after faults occur. The regular maintenance is to check, clean and calibrate the flowmeter according to a fixed time period, for example, checking the appearance of the flowmeter in a month or a quarter to check whether obvious damage, corrosion or leakage exists, calibrating once a year, comparing by using a standard flowmeter, and ensuring the accuracy of the reading. Although the normal operation of the flowmeter can be ensured to a certain extent, the method has obvious blindness, because the flowmeter can not have practical problems in a fixed period, but unnecessary maintenance is performed, and a great deal of manpower, material resources and time cost are consumed. The passive maintenance after the fault occurs is to treat the flow meter when the flow meter has failed and affects the production or monitoring work, which often leads to production interruption and serious economic loss, for example, in petrochemical industry, the flow meter failure can lead to inaccurate material conveying, affect the product quality and even cause safety accidents. In the related art, the limitation of the traditional maintenance method of the electromagnetic flowmeter is more remarkable, and the requirements of the modern industry on high reliability and stability of the flowmeter are difficult to meet. Disclosure of Invention The invention mainly aims to provide an artificial intelligence-based flow meter predictive maintenance method and system, and aims to solve the technical problem that the traditional maintenance method in the prior art is difficult to meet the requirements of modern industry on high reliability and stability of a flow meter. To achieve the above object, in a first aspect, an embodiment of the present application provides an artificial intelligence based flow meter predictive maintenance method, applied to a flow meter maintenance system, the method including: Acquiring a multi-dimensional parameter of the operation of the flowmeter, wherein the multi-dimensional parameter at least comprises one or more of a flow parameter, a conductivity parameter, a vibration parameter, an environment parameter, an instrument power supply parameter and a medium characteristic parameter; extracting characteristics of the collected multidimensional parameters to obtain static characteristics, dynamic characteristics and coupling characteristics; Inputting the extracted features into a pre-trained fault prediction model, and outputting an electrode fault prediction type and a corresponding confidence coefficient, wherein the electrode fault prediction type comprises an electrode water film covering fault, an electrode scaling covering fault, an electrode vibration loosening fault and an electrode corrosion damage fault; Performing capacitance verification on the electrode fault prediction type according to the electrode ideal capacitance value and the actual equivalent capacitance value corresponding to the electrode fault prediction type to obtain an electrode target fault type and a target confidence coefficient; According to the target fault type and the corresponding target confidence coefficient, a preset fault degradation model is called to simulate an evolution path of the fault under the current actual working condition, and the residual maintenance time of the fault is calculated; and matching and generating a targeted maintenance strategy from a preset strategy library according to the residual maintenance time of the fault and the target fault type. In one possible implementation manner, the static characteristic is a degree to which the parameter deviates from a preset normal range, including a degree to which the parameter deviates, and the deviation calculation formula is as follows: ; Wherein, delta is the parameter deviation degree, x is the parameter real-time acquisition value, x 0 is the parameter preset standard value, x_max is the parameter allowable maximum value, x_min is the parameter allowable minimum value, and when delta is more than or equal to 20%, the characteristic of high deviation is marked. In a possible implementation manner, the dynamic characteristic is a time sequence variation trend of the parameter, including a variation rate of the parameter, the variation rate is obtained by calculating a parameter variation slope in an ad