CN-121993717-A - Industrial lubricating oil online intelligent management method

CN121993717ACN 121993717 ACN121993717 ACN 121993717ACN-121993717-A

Abstract

The invention provides an online intelligent management method for industrial lubricating oil, which belongs to the technical field of industrial lubricating oil management, and comprises the steps of establishing an online monitoring module to detect viscosity, moisture content, water activity, dielectric constant, density, cleanliness level and iron spectrum image in real time, comparing a detection value with a preset threshold value to trigger early warning, adopting a graph shortest path fault tracing algorithm to locate a fault source, calling a corresponding purification module combination according to an early warning type to conduct targeted treatment, continuously monitoring until the cleaning process is recovered to be normal, and utilizing an artificial intelligent wear pattern recognition model to conduct real-time analysis and evaluation on equipment wear state on the iron spectrum image, so that the technical problems that the fault source is difficult to be quickly located and targeted purification treatment is implemented when the online monitoring parameters of the lubricating oil are abnormal are solved.

Inventors

GAO JICHAO
FENG RANRAN
Jiao Mengchao
ZHANG LUNING

Assignees

盈普(青岛)智能科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260210

Claims (10)

1. An online intelligent management method for industrial lubricating oil is characterized by comprising the following steps: An online monitoring module is established, and viscosity, moisture content, water activity, dielectric constant, density, cleanliness grade and a ferrograph image of oil in the lubricating system are detected in real time to obtain a viscosity detection value, a moisture content detection value, a water activity detection value, a dielectric constant detection value, a density detection value, a cleanliness grade detection value, an abrasion type identification result, abrasive particle size parameters and an abrasive particle concentration estimation value; Respectively comparing the viscosity detection value, the water content detection value, the water activity detection value, the dielectric constant detection value, the density detection value and the cleanliness class detection value with corresponding preset thresholds to trigger corresponding early warning; Calling an online purification module according to the early warning type to conduct targeted treatment; continuously monitoring the oil according to the set monitoring frequency in the purifying process, and removing early warning when all the detecting values in the purifying process are restored to be within the corresponding preset threshold range to generate monitoring and purifying records; Carrying out fault source positioning on the abnormal monitoring parameters by adopting a graph theory shortest path fault tracing algorithm to obtain a fault source positioning result; And periodically collecting the ferrograph image, inputting the ferrograph image into a wear pattern recognition model for equipment wear state evaluation, and generating maintenance advice when early wear, severe wear or abnormal wear is recognized.
2. The online intelligent management method of industrial lubricating oil according to claim 1, wherein the detection of the moisture content adopts a multi-frequency excitation dielectric spectrum analysis technology, and the analysis of the ferrograph image adopts a wear pattern recognition model.
3. The online intelligent management method of industrial lubricating oil according to claim 2, wherein the multi-frequency excitation dielectric spectrum analysis technology is characterized in that alternating-current excitation voltage is applied to oil under characteristic frequency points, capacitance values and loss factors are collected, a frequency-dielectric constant response curve is established, a debye relaxation model is utilized to extract water molecule relaxation peaks, and a temperature compensation model is combined to calculate a water content detection value.
4. The online intelligent management method of industrial lubricating oil according to claim 3, wherein the structure of the abrasion pattern recognition model is a fusion structure of a multitasking convolutional neural network and capsule network abrasion pattern recognition, a shared convolutional layer extracts bottom texture features and edge features, an output feature map inputs a feature pyramid structure to perform multiscale information fusion, a fusion feature map is generated, and an abrasive particle classification branch and an abrasive particle size regression branch are input.
5. The online intelligent management method of industrial lubricating oil according to claim 4, wherein the corresponding early warning is triggered, specifically the abnormal viscosity early warning is triggered when the viscosity detection value deviates from the set proportion of the viscosity value of the new oil, the water pollution early warning is triggered when the water content detection value exceeds the water content preset threshold value, and the hydrolysis risk early warning is triggered when the water activity detection value exceeds the water activity preset threshold value.
6. The method for online intelligent management of industrial lubricating oil according to claim 5, wherein the step of triggering the corresponding early warning further comprises the step of triggering the polarity pollution early warning when the dielectric constant detection value is higher than the new oil dielectric constant baseline value by a set proportion, triggering the density abnormality early warning when the density detection value is higher than the new oil density baseline value by a set proportion, and triggering the particle pollution early warning when the cleanliness class detection value is higher than the set progression.
7. The online intelligent management method of industrial lubricating oil according to claim 6, wherein the online purification module is called according to the early warning type to conduct targeted treatment, specifically, the equipment operation temperature and the equipment operation load are checked during viscosity abnormality early warning, and the combination of a coalescing filter element, a separating filter element and a water absorption filter element is adopted during water pollution early warning to conduct dehydration treatment.
8. The online intelligent management method of industrial lubricating oil according to claim 7, wherein the online intelligent management method is characterized in that an online purification module is called for targeted treatment according to an early warning type, and the online intelligent management method further comprises the steps of judging whether a water content detection value exceeds a water content preset threshold value in hydrolysis risk early warning, adopting a mesoporous adsorption filter element for treatment when the water content detection value does not exceed the water content preset threshold value, and adopting the mesoporous adsorption filter element for treatment after dehydration treatment when the water content detection value exceeds the water content preset threshold value.
9. The online intelligent management method of industrial lubricating oil according to claim 8, wherein the online intelligent management method is characterized in that an online purification module is called for targeted treatment according to an early warning type, and further comprises the steps of adopting a mesoporous adsorption filter element and a membrane hole filter element combination for purification treatment during polar pollution early warning, and adopting a deep fiber filter element and a membrane hole filter element combination for filtration purification during particle pollution early warning.
10. The online intelligent management method of industrial lubricating oil according to claim 9, wherein the oil is continuously monitored according to a set monitoring frequency in the purifying process, specifically, a purifying process viscosity detection value, a purifying process moisture content detection value, a purifying process water activity detection value, a purifying process dielectric constant detection value, a purifying process density detection value and a purifying process cleanliness class detection value are obtained, and each purifying process detection value is compared with a corresponding preset threshold value.

Description

Industrial lubricating oil online intelligent management method Technical Field The invention belongs to the technical field of industrial lubricating oil management, and particularly relates to an online intelligent management method for industrial lubricating oil. Background The industrial lubricating oil on-line monitoring technology evaluates the running state of a lubricating system by detecting parameters such as oil viscosity, moisture content, cleanliness and the like in real time. The traditional monitoring method adopts a single sensor to independently detect various indexes, triggers an alarm through a fixed threshold, and a maintainer judges the fault cause and selects a purifying measure according to experience, so that the limitations of insufficient correlation analysis among monitoring parameters, dependence on manual experience on fault source positioning and poor pertinence of purifying treatment exist. In the prior art, when a plurality of monitoring parameters are abnormal at the same time, due to the lack of a systematic fault tracing mechanism, maintainers are difficult to quickly identify a root fault source causing the abnormality, and a general purification process is often adopted to treat all abnormal conditions, so that the purification efficiency is low and the treatment period is prolonged. Meanwhile, the traditional ferrograph analysis relies on an artificial microscope to observe and judge the morphology and the abrasion type of abrasive particles, and has the problems of low efficiency, strong subjectivity and incapability of realizing real-time online analysis. That is, the prior art has the technical problems that when the online monitoring parameters of the lubricating oil are abnormal, the fault source is difficult to rapidly locate and the targeted purification treatment is performed. Disclosure of Invention In view of the above, the invention provides an online intelligent management method for industrial lubricating oil, which can solve the technical problems that in the prior art, when the online monitoring parameters of the lubricating oil are abnormal, a fault source is difficult to rapidly locate and targeted purification treatment is implemented. The invention is realized in such a way that the invention provides an online intelligent management method for industrial lubricating oil, which comprises the following steps: An online monitoring module is established, and viscosity, moisture content, water activity, dielectric constant, density, cleanliness grade and a ferrograph image of oil in the lubricating system are detected in real time to obtain a viscosity detection value, a moisture content detection value, a water activity detection value, a dielectric constant detection value, a density detection value, a cleanliness grade detection value, an abrasion type identification result, abrasive particle size parameters and an abrasive particle concentration estimation value; Respectively comparing the viscosity detection value, the water content detection value, the water activity detection value, the dielectric constant detection value, the density detection value and the cleanliness class detection value with corresponding preset thresholds to trigger corresponding early warning; Calling an online purification module according to the early warning type to conduct targeted treatment; continuously monitoring the oil according to the set monitoring frequency in the purifying process, and removing early warning when all the detecting values in the purifying process are restored to be within the corresponding preset threshold range to generate monitoring and purifying records; Carrying out fault source positioning on the abnormal monitoring parameters by adopting a graph theory shortest path fault tracing algorithm to obtain a fault source positioning result; And periodically collecting the ferrograph image, inputting the ferrograph image into a wear pattern recognition model for equipment wear state evaluation, and generating maintenance advice when early wear, severe wear or abnormal wear is recognized. The method comprises the steps of detecting moisture content, adopting a multi-frequency excitation dielectric spectrum analysis technology, and analyzing a ferrograph image by adopting a wear pattern recognition model. The multi-frequency excitation dielectric spectrum analysis technology specifically comprises the steps of applying alternating current excitation voltage to oil liquid at characteristic frequency points, collecting capacitance values and loss factors, establishing a frequency-dielectric constant response curve, extracting water molecule relaxation peaks by using a Debye relaxation model, and calculating a water content detection value by combining a temperature compensation model. The structure of the abrasion pattern recognition model is a fusion structure of a multitasking convolutional neural network and capsule network abrasion pattern recognit