CN-121997030-A - Dynamic analysis and early warning method for maintenance data of ship equipment

CN121997030ACN 121997030 ACN121997030 ACN 121997030ACN-121997030-A

Abstract

The invention provides a dynamic analysis and early warning method for maintenance data of ship equipment, which relates to the technical field of data processing, and comprises the following steps of 1, processing multi-dimensional operation state time sequence data of the ship equipment, constructing a unified time sequence data set, and extracting health feature vectors reflecting the mechanical performance and the overall operation state of the equipment; aiming at a key rod member structure in the equipment, based on axial load data and rod member geometric parameters monitored in real time, calculating axial stress, combining material allowable stress, and comparing to obtain an axial tensile strength safety coefficient to form a health feature vector. The invention realizes the advanced identification and intelligent early warning of the hidden trouble, improves the maintenance efficiency and ensures the operation safety of the ship.

Inventors

WU BINGKUN
YAO FENG
WANG ZHENG
HUANG SHIYONG
WANG PENGCHENG

Assignees

众数(厦门)信息科技有限公司
众数智能(厦门)科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260317

Claims (10)

1. The method for dynamically analyzing and early warning the maintenance data of the ship equipment is characterized by comprising the following steps: Step 1, processing multi-dimensional operation state time sequence data of ship equipment, constructing a unified time sequence data set, extracting health feature vectors reflecting the mechanical performance and the overall operation state of the equipment, calculating axial stress based on axial load data and rod geometric parameters monitored in real time aiming at key rod structures in the equipment, and combining material allowable stress to obtain an axial tensile strength safety coefficient through comparison to form the health feature vectors; step 2, inputting the health feature vector into a pre-trained deep learning ship equipment state prediction model to obtain a health degree evaluation result and an abnormal probability prediction value; Step 3, generating dynamic correction coefficients and correcting the health evaluation result and the abnormal probability predicted value through multi-parameter association analysis based on real-time data of three core monitoring indexes in the unified time sequence data set to obtain the corrected health evaluation result and the corrected abnormal probability predicted value; step 4, based on the corrected health evaluation result and the abnormal probability predicted value, carrying out real-time early warning judgment by combining a dynamic self-adaptive threshold judgment strategy; And step 5, starting a maintenance coordination and management response flow according to the alarm information, integrating the alarm information into a ship maintenance management platform, generating a maintenance work order, recommending a maintenance time window and a personnel configuration scheme to form a maintenance execution record, and simultaneously combining the existing early warning record and maintenance feedback data to dynamically optimize a maintenance plan to form a closed-loop maintenance business management flow.
2. The method for dynamically analyzing and pre-warning the maintenance data of the ship equipment according to claim 1, wherein the health feature vector comprises a root mean square value of a vibration signal, spectrum dominant frequency energy and change trend features of temperature and pressure.
3. The method for dynamically analyzing and pre-warning the maintenance data of the ship equipment according to claim 2, wherein the steps of processing the multi-dimensional operation state time sequence data of the ship equipment, constructing a unified time sequence data set, extracting health feature vectors reflecting the mechanical properties and the overall operation state of the equipment, and comprising: Carrying out data cleaning on multi-dimensional operation state time sequence data acquired from different sensors, eliminating abnormal values generated by noise or transmission errors, and filling the missing data by adopting linear interpolation to obtain cleaned time sequence data; resampling according to the cleaned time sequence data and the uniform time frequency to realize time alignment of multi-source data and form a structured uniform time sequence data set; Extracting characteristic parameters capable of representing mechanical performance and overall operation state of equipment according to the unified time sequence data set to obtain an initial characteristic set; And screening key features by adopting feature importance evaluation based on random forests according to the initial feature set to form a health feature vector.
4. The method for dynamically analyzing and pre-warning the maintenance data of the ship equipment according to claim 3, wherein the characteristic parameters comprise statistical characteristics reflecting data distribution characteristics, frequency domain characteristics reflecting signal frequency composition and time-frequency characteristics reflecting signal time-frequency local characteristics.
5. The method for dynamically analyzing and pre-warning maintenance data of ship equipment according to claim 4, wherein for a key rod structure in the equipment, based on real-time monitored axial load data and rod geometric parameters, calculating axial stress, and combining material allowable stress, and obtaining an axial tensile strength safety coefficient through comparison, to form a health feature vector, comprising: acquiring axial load data born by the rod in real time through a strain sensor or a load sensor arranged on the key rod, and acquiring geometric parameters and material identifiers of the rod from an equipment design parameter database, wherein the geometric parameters at least comprise cross sectional area, length and section moment of inertia; based on the axial load data and the cross-sectional area, calculating to obtain an axial stress value of the rod piece under the current working condition, and according to the material identification, calling an allowable stress value of a corresponding material from a preset material attribute database; and taking the axial tensile strength safety coefficient as one characteristic component in the health characteristic vector, and combining the axial tensile strength safety coefficient with the geometric parameter and the associated monitoring data together to form the health characteristic vector capable of comprehensively representing the structural strength state of the key rod piece.
6. The method for dynamically analyzing and pre-warning the maintenance data of the ship equipment according to claim 5, wherein the step of inputting the health feature vector into the pre-trained state prediction model of the deep learning ship equipment to obtain the health evaluation result and the abnormal probability prediction value comprises the following steps: Collecting operation data of ship equipment under a past working condition as sample operation data, preprocessing the sample operation data, extracting multi-dimensional health features to form sample health feature vectors, and simultaneously obtaining equipment state labels corresponding to the sample health feature vectors to construct a training sample set; Based on a training sample set, a deep learning ship equipment state prediction model is constructed, a sample health feature vector is used as input, equipment health state labels at corresponding moments are used as output for training, and parameters of the deep learning ship equipment state prediction model are optimized by minimizing prediction errors, so that a trained ship equipment state prediction model is obtained; Collecting current operation data of ship equipment in real time, preprocessing the current operation data and extracting features to generate a health feature vector at the current moment; and inputting the health feature vector at the current moment into a trained ship equipment state prediction model, and obtaining equipment health degree score and abnormal probability value at the current moment through forward propagation calculation.
7. The method of claim 6, wherein the health score is used to quantify the degree of equipment performance degradation and the anomaly probability value is used to characterize the likelihood of equipment failure.
8. The method for dynamically analyzing and pre-warning maintenance data of ship equipment according to claim 7, wherein generating a dynamic correction coefficient and correcting a health evaluation result and an abnormal probability prediction value by multi-parameter association analysis based on real-time data of three core monitoring indexes in a unified time sequence data set to obtain the corrected health evaluation result and the corrected abnormal probability prediction value comprises: Extracting current time data of three core monitoring indexes of vibration amplitude, temperature value and pressure value and continuous data in a preset time window from the unified time sequence data set in real time to obtain a real-time monitoring data set; Based on the real-time monitoring data set, calculating real-time correlation coefficients of vibration amplitude, temperature value and pressure value at the current moment to generate a real-time correlation matrix, and simultaneously, calling the reference correlation coefficient of the equipment in a corresponding time window under a standard working condition or a fault-free running state from a pre-stored reference working condition feature library to generate a reference correlation matrix; Element-by-element comparison is carried out on the real-time correlation matrix and the reference correlation matrix, deviation values of the corresponding correlation coefficients are calculated, and the deviation values are synthesized to obtain a correlation deviation value capable of quantitatively reflecting the deviation degree of the current working condition relative to the standard working condition; Dynamically generating a correction coefficient according to the association deviation amount through a preset mapping rule, wherein when the association deviation amount exceeds a preset silence threshold range, the condition is judged to change and a nonlinear correction coefficient is obtained; And (3) applying the correction coefficient to the initial health degree evaluation result and the abnormal probability prediction value, and correcting the original prediction value by adopting weighted summation to finally obtain a corrected health degree evaluation result and an abnormal probability prediction value which are more in line with the current actual physical state of the equipment.
9. The method for dynamically analyzing and pre-warning the maintenance data of the ship equipment according to claim 8, wherein the method is characterized in that based on the corrected health evaluation result and the abnormal probability predicted value, real-time pre-warning judgment is performed by combining a dynamic self-adaptive threshold judgment strategy, and when the predicted value exceeds the dynamic pre-warning threshold, alarm information is generated, comprising: based on the corrected health degree evaluation result and the abnormal probability prediction value, combining the health degree evaluation result serving as the latest data point with the corrected health degree evaluation result at the previous moment stored in the database to form a health degree evaluation time sequence comprising the current state and the past state; Selecting corrected health degree evaluation results of the latest N moments from the health degree evaluation time sequence, calculating an arithmetic average value of the N results to be used as a health degree reference value of the current moment, and calculating a standard deviation of the N results to be used as a health degree fluctuation range of the current moment; Dynamically generating an abnormal probability early warning threshold value at the current moment according to a preset conversion rule according to the health standard value and the health fluctuation range, so that the early warning threshold value adaptively follows the normal fading trend and the working condition change of the equipment; Judging whether a preset alarm triggering condition is met according to the comparison result, wherein the corrected abnormal probability predicted value continuously exceeds the dynamic early warning threshold for a plurality of times or exceeds the dynamic early warning threshold for a single time and exceeds the preset range; if the alarm triggering condition is judged to be met, generating alarm information, wherein the alarm information comprises equipment identification, abnormal occurrence time, abnormal probability value, related monitoring index data and recommended maintenance level.
10. The method for dynamically analyzing and pre-warning the maintenance data of the ship equipment according to claim 9, wherein the method for dynamically optimizing the maintenance plan and forming the closed-loop maintenance business management flow by combining the existing pre-warning record and the maintenance feedback data comprises the steps of: Automatically pushing the alarm information to a ship maintenance management platform, extracting the equipment type, the fault mode and the abnormal occurrence time, and generating a maintenance work order comprising detailed task description by combining current navigation state data of the ship; Based on a maintenance work order, acquiring a ship navigation plan, the current availability status of equipment and stored past maintenance data, and calculating and recommending a final maintenance time window through a preset optimization algorithm; According to the technical requirements of maintenance tasks in a maintenance work order and a recommended maintenance time window, calling a maintenance personnel skill level database and personnel availability information, automatically matching personnel with corresponding skill levels available in the maintenance time window, and generating a recommended personnel configuration scheme; After the maintenance operation is actually executed according to the recommended personnel configuration scheme, acquiring actual operation content, time consumption and spare part replacement information of the maintenance, and inputting the actual operation content, time consumption and spare part replacement information into a database of a ship maintenance management platform to form a complete maintenance execution record associated with a maintenance worksheet; The existing alarm records and maintenance execution records stored in the ship maintenance management platform are collected regularly, maintenance rules and optimizable links are identified through statistical analysis, and maintenance periods, early warning threshold setting and resource allocation strategies in a maintenance plan are dynamically adjusted according to statistical analysis results, so that closed-loop maintenance business management optimization is realized.

Description

Dynamic analysis and early warning method for maintenance data of ship equipment Technical Field The invention relates to the technical field of data processing, in particular to a dynamic analysis and early warning method for maintenance data of ship equipment. Background In the development process of intelligent operation and maintenance management of ships, the assessment and early warning of the operation state of key mechanical equipment are important bases for guaranteeing navigation safety and reducing maintenance cost, and at present, the health management of the ship equipment mostly depends on a preset single threshold value alarm or a regular offline maintenance mechanism. However, in practical applications, there are some limitations in the existing monitoring and analyzing methods, for example, in the scenario that key rods in a ship propulsion system, such as a middle shaft and a thrust shaft, are taken as monitoring objects, in most of the existing methods, axial load data is obtained through a strain sensor arranged on the shaft system, axial stress is calculated according to the axial load data, and then the axial stress is simply compared with allowable stress of materials to determine whether a strength safety coefficient is within an allowable range, although the method can reflect static stress conditions of the rods at a specific moment, the method is difficult to effectively cope with severe fluctuation of dynamic loads under complex sea conditions, due to coupling actions of sea wave impact, propeller exciting force and hull deformation, a great amount of non-stationary noise and transient impact signals are often mixed in the axial load data, if the safety coefficient is calculated directly based on the non-deeply cleaned single-point data, and mapped to a health feature vector of the whole equipment based on the safety coefficient, and a deviation may be caused in the evaluation result, so that continuous trend degradation of mechanical performance of the equipment is difficult to accurately reflect. Disclosure of Invention The technical problem to be solved by the invention is to provide a dynamic analysis and early warning method for maintenance data of ship equipment, so that the early identification and intelligent early warning of fault hidden danger are realized, the maintenance efficiency is improved, and the operation safety of the ship is ensured. In order to solve the technical problems, the technical scheme of the invention is as follows: in a first aspect, a method for dynamically analyzing and pre-warning maintenance data of ship equipment, the method comprising: Step 1, processing multi-dimensional operation state time sequence data of ship equipment, constructing a unified time sequence data set, extracting health feature vectors reflecting the mechanical performance and the overall operation state of the equipment, calculating axial stress based on axial load data and rod geometric parameters monitored in real time aiming at key rod structures in the equipment, and combining material allowable stress to obtain an axial tensile strength safety coefficient through comparison to form the health feature vectors; step 2, inputting the health feature vector into a pre-trained deep learning ship equipment state prediction model to obtain a health degree evaluation result and an abnormal probability prediction value; Step 3, generating dynamic correction coefficients and correcting the health evaluation result and the abnormal probability predicted value through multi-parameter association analysis based on real-time data of three core monitoring indexes in the unified time sequence data set to obtain the corrected health evaluation result and the corrected abnormal probability predicted value; step 4, based on the corrected health evaluation result and the abnormal probability predicted value, carrying out real-time early warning judgment by combining a dynamic self-adaptive threshold judgment strategy; And step 5, starting a maintenance coordination and management response flow according to the alarm information, integrating the alarm information into a ship maintenance management platform, generating a maintenance work order, recommending a maintenance time window and a personnel configuration scheme to form a maintenance execution record, and simultaneously combining the existing early warning record and maintenance feedback data to dynamically optimize a maintenance plan to form a closed-loop maintenance business management flow. The scheme of the invention at least comprises the following beneficial effects: The method comprises the steps of uniformly processing multidimensional time sequence data of ship equipment and extracting health feature vectors, realizing comprehensive quantitative characterization of the mechanical performance and the overall operation state of the equipment, simultaneously carrying out axial stress and strength safety coefficient calculation