CN-121980464-A - Method and system for diagnosing faults of edge equipment in complex port environment

CN121980464ACN 121980464 ACN121980464 ACN 121980464ACN-121980464-A

Abstract

The invention discloses a fault diagnosis method and a fault diagnosis system for edge equipment in a complex port environment, which relate to the technical field of fault detection and comprise the steps of continuously collecting temperature, vibration and noise multi-source heterogeneous data, and obtaining a dimensionality reduction data set through light preprocessing and dimensionality reduction processing; and constructing a light deep learning model based on integrated learning, fusing multi-source dimension reduction data and the three-level coefficient to perform localized real-time reasoning, and outputting a fault mode. The method comprises the steps of optimizing an anti-interference data acquisition device and a robustness algorithm, effectively resisting the influence of strong electromagnetic noise and vibration interference on a diagnosis result, improving the diagnosis accuracy in a complex environment, constructing a multi-source heterogeneous data embedded feature fusion method based on a device damage mechanism, fully utilizing the complementary information of temperature, vibration and noise data, and improving the comprehensiveness and the accuracy of fault diagnosis through three-level fault correlation analysis.

Inventors

LIU LEILEI
LIU JIAN
LI XIANRUI
XU BIN
ZHAO HAOXU
Zhou Sanbo
LI BINGSHUAI
WANG ZHE
ZHANG LEI
LI ZHANGYUN

Assignees

交通运输部天津水运工程科学研究所

Dates

Publication Date: 20260505
Application Date: 20260203

Claims (10)

1. The method for diagnosing the fault of the edge equipment in the complex port environment is characterized by comprising the following steps of: acquiring temperature information, vibration information and current information of edge equipment to obtain a temperature information set, a vibration information set and a current information set; Performing dimension reduction processing on the temperature information set, the vibration information set and the current information set respectively to obtain a dimension reduction temperature information set, a dimension reduction vibration information set and a dimension reduction current information set; performing correlation analysis on the dimension reduction temperature information set, the dimension reduction vibration information set and the dimension reduction current information set to obtain a plurality of correlation analysis coefficients; And constructing a lightweight deep learning model, respectively inputting a dimension reduction temperature information set, a dimension reduction vibration information set and a dimension reduction current information set into corresponding fault identification branches in the model, determining the reasoning weight of each fault identification branch according to a plurality of correlation analysis coefficients, and carrying out multi-source data fusion reasoning by an embedded feature fusion method to obtain a fault diagnosis result.
2. The method for diagnosing an edge device fault in a complex port environment according to claim 1, wherein the step of collecting temperature information of the edge device comprises: When the port edge equipment is started, acquiring temperature information of key positions on the edge equipment in a plurality of time nodes to obtain a temperature array set; aiming at temperature data abnormality caused by strong electromagnetic interference of ports, carrying out outlier rejection and data calibration on a temperature array of each time node; and carrying out temperature interpolation processing on all key areas on the edge equipment according to the calibrated temperature array to obtain a temperature information set.
3. The method for diagnosing an edge device fault in a complex port environment according to claim 1, wherein the dimension reduction process comprises: calculating to obtain a reference core temperature according to the statistical characteristics of the temperature information set; Calculating distribution probability according to temperature difference values of other key temperatures and reference core temperatures in the temperature information set and by combining with a temperature change rule of equipment in a port environment, and obtaining key probability distribution; Adopting a lightweight dimension reduction algorithm adapting to the edge equipment, and carrying out dimension reduction processing on the temperature information set according to the key probability distribution to obtain a dimension reduction temperature information set; And acquiring time nodes of a plurality of dimension reduction key temperature information in the dimension reduction temperature information set, extracting multi-source data of corresponding time nodes in the vibration information set and the current information set, and extracting features strongly related to temperature change by an embedded feature fusion method to obtain the dimension reduction vibration information set and the dimension reduction current information set.
4. The method for diagnosing faults of edge equipment in a complex port environment according to claim 3, wherein the step of adopting a lightweight dimension reduction algorithm of adaptive edge equipment to perform dimension reduction processing on the temperature information set according to key probability distribution to obtain the dimension reduction temperature information set comprises the following steps: Setting a preset dimension reduction ratio and a preset dimension reduction frequency which adapt to the computing capacity of the edge equipment; Randomly extracting key temperatures from the temperature information set according to a preset dimension reduction proportion to obtain a first dimension reduction temperature information set; Calculating distribution probability according to temperature difference values of a plurality of first dimension reduction key temperatures and the reference core temperature in the first dimension reduction temperature information set to obtain first dimension reduction key probability distribution; Calculating the similarity between the first dimensionality reduction key probability distribution and the key probability distribution to obtain a first key dimensionality reduction; continuously randomly extracting dimension reduction according to a preset dimension reduction proportion to obtain a second dimension reduction temperature information set, and processing to obtain a second key dimension reduction; and repeating the random extraction dimension reduction processing until the preset dimension reduction times are reached, and outputting a dimension reduction temperature information set which has the maximum key dimension reduction and meets the calculation delay requirement of the edge equipment.
5. The method for diagnosing faults of edge equipment in a complex harbor environment according to claim 1, wherein the performing correlation analysis on the dimension-reduced temperature information set, the dimension-reduced vibration information set and the dimension-reduced current information set respectively two by two to obtain a plurality of correlation analysis coefficients comprises: performing primary fault correlation analysis based on a device thermal damage mechanism according to the dimension reduction temperature information set and the dimension reduction current information set to obtain a first coefficient; adopting the dimension-reduction vibration information set and the dimension-reduction temperature information set to perform secondary fault correlation analysis based on an equipment damage mechanism to obtain a second coefficient; And adopting the dimension-reduced current information set and the dimension-reduced vibration information set to perform three-level fault correlation analysis aiming at a strong electromagnetic noise interference scene so as to obtain a third coefficient.
6. The method for diagnosing an edge device in a complex harbor environment according to claim 5, wherein the performing a first-level fault correlation analysis according to the reduced-dimension temperature information set and the reduced-dimension current information set, and obtaining the first coefficient comprises: Deleting the initial unstable stage data in the dimension reduction temperature information set, and retaining the effective data after the equipment stably operates; and combining the dimension reduction current information set, performing primary fault correlation analysis based on a device thermal damage mechanism, and calculating by adopting a pearson correlation coefficient algorithm to obtain a first coefficient: ; wherein r is a first coefficient, m is the number of effective data in the dimension-reduction temperature information set and the dimension-reduction current information set, For the ith dimension reduction key temperature in the dimension reduction temperature information set, For the i-th part deformation parameter in the dimension-reduction current information set, And The average values of the dimension reduction temperature information set and the dimension reduction current information set are respectively.
7. The method for diagnosing an edge device failure in a complex harbor environment according to claim 5, wherein said constructing a lightweight deep learning model comprises: collecting fault diagnosis historical data of edge equipment in a complex harbor environment, and obtaining a sample dimension-reduction temperature information set, a sample dimension-reduction vibration information set, a sample dimension-reduction current information set and a corresponding sample fault mode set, wherein the sample information comprises labeling data under strong electromagnetic noise and vibration interference scenes; the sample dimension-reduction temperature information set, the sample dimension-reduction vibration information set and the sample dimension-reduction current information set are respectively adopted as inputs, the sample fault mode set is adopted as an output, and a temperature fault identification branch, a vibration fault identification branch and a noise fault identification branch are constructed based on an integrated learning algorithm; each fault identification branch adopts a lightweight network structure, and the running efficiency of the model on the edge equipment is improved through a quantitative compression optimization algorithm.
8. The method for diagnosing an edge device fault in a complex port environment according to claim 7, wherein the fault identification branch comprises a plurality of fault identification paths, each path being optimized for different types of damage faults; and integrating four fault recognition branches, embedding a multi-source heterogeneous data feature fusion module, and obtaining a lightweight deep learning model which supports localized real-time reasoning of the edge equipment.
9. The method for diagnosing the fault of the edge equipment in the complex harbor environment according to claim 8, wherein the step of performing multi-source data fusion reasoning through the embedded feature fusion method to obtain the fault diagnosis result comprises the following steps: Determining a first inference weight, a second inference weight and a third inference weight according to the first coefficient, the second coefficient and the third coefficient; The dimension-reduction temperature information set, the dimension-reduction vibration information set and the dimension-reduction current information set are respectively input into the temperature fault identification branch, the vibration fault identification branch and the noise fault identification branch, calculation resources are allocated according to corresponding reasoning weights for fault analysis, and a plurality of basic fault modes are output; and fusing all the basic fault modes through an embedded feature fusion module, screening the basic fault mode which has the highest occurrence frequency and accords with the equipment damage mechanism, and taking the basic fault mode as a final fault mode, wherein the final fault mode comprises a fault type, a damage degree, a fault position and a maintenance suggestion.
10. An edge equipment fault diagnosis system in a complex port environment, comprising: The acquisition module is used for acquiring temperature information, vibration information and current information of the edge equipment and obtaining a temperature information set, a vibration information set and a current information set; The dimension reduction module is used for respectively carrying out dimension reduction processing on the temperature information set, the vibration information set and the current information set to obtain a dimension reduction temperature information set, a dimension reduction vibration information set and a dimension reduction current information set; The correlation analysis module is used for carrying out correlation analysis on the dimension reduction temperature information set, the dimension reduction vibration information set and the dimension reduction current information set to obtain a plurality of correlation analysis coefficients; The reasoning module is used for constructing a lightweight deep learning model, respectively inputting the dimension-reduction temperature information set, the dimension-reduction vibration information set and the dimension-reduction current information set into corresponding fault recognition branches in the model, and determining the reasoning weight of each fault recognition branch according to a plurality of correlation analysis coefficients; the fault diagnosis module is used for carrying out multi-source data fusion reasoning through an embedded feature fusion method to obtain a fault diagnosis result.

Description

Method and system for diagnosing faults of edge equipment in complex port environment Technical Field The invention relates to the technical field of fault detection, in particular to a method and a system for diagnosing faults of edge equipment in a complex port environment. Background The port edge equipment is a core infrastructure for port logistics operation, and the operation state of the port edge equipment directly affects the port operation efficiency and safety. The port environment has the complex characteristics of strong electromagnetic noise, severe vibration interference, large temperature and humidity change and the like, so that the fault of the edge equipment presents diversity, complexity and burst. The prior edge equipment fault diagnosis technology mainly has the following problems: The method relies on single sensor data or simple parameter monitoring, complementary information of multi-source heterogeneous data is not fully utilized, diagnosis accuracy is low, algorithm robustness is insufficient, data interference is easy to occur under strong electromagnetic noise and vibration interference scenes of a port, diagnosis false alarm rate and false alarm rate are high, the existing deep learning model is high in calculation complexity, limited calculation resources of edge equipment cannot be adapted, localized real-time reasoning is difficult to achieve, fault correlation analysis aiming at a port edge equipment damage mechanism is lacked, and matching degree of fault diagnosis and actual operation state of equipment is low. Therefore, how to provide a method and a system for diagnosing faults of edge equipment in a complex port environment, which overcome the defects in the prior art, are the problems to be solved by those skilled in the art. Disclosure of Invention In view of the above, the invention provides a method and a system for diagnosing faults of edge equipment in a complex port environment, which solve the problems that the fault diagnosis of the edge equipment in the complex port environment in the prior art is low in accuracy, insufficient in robustness and incapable of localization real-time reasoning. In order to achieve the above purpose, the present invention adopts the following technical scheme: a fault diagnosis method for edge equipment in a complex port environment comprises the following steps: acquiring temperature information, vibration information and current information of edge equipment to obtain a temperature information set, a vibration information set and a current information set; Performing dimension reduction processing on the temperature information set, the vibration information set and the current information set respectively to obtain a dimension reduction temperature information set, a dimension reduction vibration information set and a dimension reduction current information set; performing correlation analysis on the dimension reduction temperature information set, the dimension reduction vibration information set and the dimension reduction current information set to obtain a plurality of correlation analysis coefficients; And constructing a lightweight deep learning model, respectively inputting a dimension reduction temperature information set, a dimension reduction vibration information set and a dimension reduction current information set into corresponding fault identification branches in the model, determining the reasoning weight of each fault identification branch according to a plurality of correlation analysis coefficients, and carrying out multi-source data fusion reasoning by an embedded feature fusion method to obtain a fault diagnosis result. Optionally, the collecting temperature information of the edge device includes: When the port edge equipment is started, acquiring temperature information of key positions on the edge equipment in a plurality of time nodes to obtain a temperature array set; aiming at temperature data abnormality caused by strong electromagnetic interference of ports, carrying out outlier rejection and data calibration on a temperature array of each time node; and carrying out temperature interpolation processing on all key areas on the edge equipment according to the calibrated temperature array to obtain a temperature information set. Optionally, the dimension reduction processing includes: calculating to obtain a reference core temperature according to the statistical characteristics of the temperature information set; Calculating distribution probability according to temperature difference values of other key temperatures and reference core temperatures in the temperature information set and by combining with a temperature change rule of equipment in a port environment, and obtaining key probability distribution; Adopting a lightweight dimension reduction algorithm adapting to the edge equipment, and carrying out dimension reduction processing on the temperature information set according to the key probab