Search

CN-122020465-A - Intelligent monitoring and evaluating system for metal health state of generator set

CN122020465ACN 122020465 ACN122020465 ACN 122020465ACN-122020465-A

Abstract

The invention discloses an intelligent monitoring and evaluating system for the metal health state of a generator set, and belongs to the technical field of power generation equipment state monitoring. The system comprises a data acquisition and preprocessing module, a data fusion and feature extraction module, a metal health state evaluation module, a residual life prediction and fault tracing module and a visual intelligent decision support module. The method comprises the steps of extracting cross-scale damage characteristics by fusing macroscopic operation data and microscopic nondestructive detection signals, calculating comprehensive health indexes by adopting a self-adaptive weighting fusion mode of a physical mechanism model and a data driving model, further realizing dynamic residual life prediction and failure cause intelligent tracing, and finally providing grading early warning and maintenance decision advice through a three-dimensional visual interface. The invention solves the problems of single monitoring dimension, isolated model and static prediction in the prior art, and realizes the omnibearing, high-precision, interpretable evaluation and predictive maintenance support of the health state of the metal part.

Inventors

  • LIU JIAWEI

Assignees

  • 西安正卓检测技术有限公司

Dates

Publication Date
20260512
Application Date
20260128

Claims (10)

  1. 1. An intelligent monitoring and evaluating system for the metal health state of a generator set, which is characterized by comprising: The data acquisition and preprocessing module (1) is used for acquiring macroscopic operation data, microscopic nondestructive detection data and structural parameter data of the generator set in real time, and preprocessing and standardizing the data; The data fusion and feature extraction module (2) is used for carrying out space-time alignment and feature level fusion on the preprocessed multi-source heterogeneous data and extracting trans-scale features reflecting the damage mechanism of the metal part; The mechanism-data double-driven metal health state evaluation module (3) is used for constructing a damage evolution model based on a physical failure mechanism and a performance degradation model based on machine learning, and generating a comprehensive health index of the metal part through a weighted fusion mechanism; The residual life prediction and fault tracing module (4) is used for predicting the dynamic residual life of the metal part based on the comprehensive health index and the part operation load spectrum and identifying a dominant failure mode and key cause parameters which cause the health state degradation; and the visual intelligent decision support module (5) is used for carrying out multidimensional visual presentation on the evaluation and prediction results and generating hierarchical early warning information and maintenance strategy suggestions according to preset rules.
  2. 2. The system for intelligently monitoring and evaluating the metal health status of a generator set according to claim 1, wherein the data acquisition and preprocessing module (1) comprises: The macroscopic operation data acquisition unit (11) is used for acquiring operation temperature, pressure, flow, vibration signals and load spectrum data of key parts of a main steam pipeline, a reheat steam pipeline, a cylinder and a rotor of the generator set; A microscopic nondestructive testing data acquisition unit (12) integrating at least two technologies of nonlinear ultrasonic testing, barkhausen noise testing, instrumented indentation testing and magnetic memory testing and used for acquiring signals reflecting microstructure degradation, creep damage, fatigue crack initiation and thermal aging states of the metal material; the structural parameter data management unit (13) is used for storing and managing the material marks, service years, geometric dimensions, historical overhaul records and defect maps of the metal parts; and the data preprocessing unit (14) is used for carrying out noise filtering, outlier detection and restoration, missing value interpolation and time sequence alignment on the acquired original data.
  3. 3. The system for intelligently monitoring and evaluating the metal health status of a generator set according to claim 1 or 2, wherein the data fusion and feature extraction module (2) is specifically configured to: dividing the running state of the unit through a working condition identification model, and mapping monitoring data under different working conditions to standard reference working conditions; performing time axis alignment on time sequence data from sensors with different sampling frequencies by adopting a dynamic time warping algorithm; The method comprises the steps of extracting trans-scale characteristics, wherein the trans-scale characteristics comprise macro characteristics of oxide skin growth thickness, creep strain accumulation and fatigue damage degree calculated based on a mechanism model, and micro characteristics of nonlinear ultrasonic coefficients, barkhausen noise RMS values, microhardness changes and magnetic memory signal gradient K values obtained based on nondestructive testing signal analysis; constructing the joint feature vector of the trans-scale feature.
  4. 4. A generator set metal health status intelligent monitoring and assessment system according to claim 3, wherein said mechanism-data dual-driven metal health status assessment module (3) comprises: The mechanism model submodule (31) is used for constructing an oxide skin shedding risk prediction model and a creep-fatigue interactive damage evolution model which take the wall temperature and the stress of a pipeline as inputs based on material mechanics, fracture mechanics and high-temperature corrosion theory; the data model submodule (32) is used for taking the joint feature vector as input and outputting the performance degradation trend of the metal part through a trained deep confidence network or long-term and short-term memory network model; The self-adaptive weighted fusion unit (33) is used for dynamically adjusting the weights of the output results of the mechanism model submodule (31) and the data model submodule (32) according to the current working condition, the component type and the data confidence level, and calculating to obtain a comprehensive health index CHI, CHI E [0, 1], wherein 0 represents complete failure and 1 represents good condition.
  5. 5. The intelligent monitoring and evaluating system for metal health of a generator set according to claim 4, wherein the residual life prediction and fault tracing module (4) comprises: The residual life prediction unit (41) is used for inputting the time sequence of the comprehensive health index and the load spectrum currently born by the component into a particle filtering or Wiener process degradation model, and predicting the time when the health index of the component first touches the failure threshold value as a dynamic residual life; the fault tracing unit (42) is used for analyzing the contribution degree of the input features in the data model submodule (32) to the health state evaluation result based on a Shapley value or a local interpretable model agnostic interpretation method and identifying key sensitive features and corresponding physical failure modes which cause health degradation; and the risk matrix construction unit (43) is used for combining the failure probability and the failure result severity of the component to construct a dynamic risk matrix and classifying the safety risk of the metal component.
  6. 6. The system for intelligent monitoring and assessment of the metal health of a generator set according to claim 1, characterized in that said visual intelligent decision support module (5) comprises: The health-state panoramic cockpit (51) is used for displaying the real-time health index, the residual life and the risk level of each metal part in a layered and color-separated manner by taking the factory three-dimensional digital twin model as a background; The trend prediction and early warning center (52) is used for displaying a historical change curve, a future prediction track and failure probability of the health index of the key component, and triggering sound and light alarm and work order pushing of corresponding grades when the health index is lower than a yellow early warning threshold or the residual life is lower than a set value; A maintenance decision knowledge base (53) for storing maintenance recommendations generated based on historical cases, expert rules and reliability-focused maintenance strategies, including inspection cycles, recommended non-destructive inspection methods, possible repair or replacement schemes.
  7. 7. The intelligent monitoring and assessment system for metal health status of a generator set according to claim 1, wherein the system further comprises an edge-cloud co-computing architecture (6): The edge intelligent node (61) is deployed on the site of the generator set, is internally provided with a lightweight inference model and is used for carrying out real-time processing and preliminary diagnosis on vibration and temperature data acquired at high frequency so as to realize millisecond-level abnormal perception; And the cloud analysis platform (62) is used for receiving the concentrated characteristic data and the preliminary result uploaded by the edge node, running a complex mechanism-data fusion model, a life prediction model and a global optimization algorithm, and periodically transmitting updated model parameters to the edge node.
  8. 8. A method for intelligently monitoring and evaluating the metal health status of a generator set based on the system of any one of claims 1-7, comprising the following steps: step S1, macro operation data, micro nondestructive detection data and structural parameter data of a generator set are acquired and processed in a standardized manner in real time through a data acquisition and preprocessing module (1); s2, carrying out space-time alignment and feature fusion on multi-source data through a data fusion and feature extraction module (2) to extract a cross-scale feature vector; step S3, calculating a real-time comprehensive health index of the metal part by means of a mechanism-data double-drive metal health state evaluation module (3) and combining a physical mechanism with the output of a data drive model; s4, predicting the residual life of the component based on the health index and the load spectrum by a residual life prediction and fault tracing module (4), and tracing the leading failure reason; And S5, displaying the evaluation and prediction result through a visual intelligent decision support module (5), and generating an early warning and maintenance decision.
  9. 9. The method for intelligently monitoring and evaluating metal health of a generator set according to claim 8, wherein in step S3, the calculation formula of the integrated health index CHI is: CHI(t) = α(t) · MHI(t) + β(t) · DHI(t); the MHI (t) is a health index output by a mechanism model, the DHI (t) is a health index output by a data driving model, alpha (t) and beta (t) are dynamic weight coefficients, alpha (t) +beta (t) =1, and the dynamic weight coefficients are adaptively adjusted according to the current data quality, the historical accuracy of the model under the working condition and the service stage of the component.
  10. 10. A computer readable storage medium having stored thereon a computer program, wherein the program when executed by a processor implements the method for intelligently monitoring and assessing the metal health of a generator set according to claim 8 or 9.

Description

Intelligent monitoring and evaluating system for metal health state of generator set Technical Field The invention relates to the technical field of power generation equipment state monitoring, fault diagnosis and predictive maintenance, in particular to an intelligent monitoring and evaluating system for the metal health state of a generator set, and particularly relates to an intelligent monitoring and evaluating system and method for the metal health state of the generator set, which are integrated with multi-source data and integrated with physical mechanisms and artificial intelligence. Background The generator set, especially the thermal power and gas turbine set, has the key metal parts of high temperature and high pressure pipeline, such as main steam pipe, reheat section, cylinder, rotor, etc. in harsh work condition, and has the combined action of high temperature creep, fatigue load, oxidation corrosion, etc. The boiler heating surface pipes (such as an economizer, a water cooling wall, a superheater and a reheater) are exposed to high-temperature flue gas, corrosive media and complex stress environments for a long time, and the problems of typical damage such as wall thickness reduction, pipe expansion, inner wall oxide skin thickening and falling, metallographic structure aging (such as pearlite spheroidization and graphitization) are faced. Failure of these components often results in unplanned outages and even catastrophic events, resulting in significant economic losses and safety risks. Therefore, accurate monitoring, evaluation and life prediction of the health state of the metal parts of the generator set are key to realizing predictive maintenance and guaranteeing safe and stable operation of the power system. The existing power generation equipment state monitoring technology mainly has the following limitations: The monitoring dimension is single, the association of microcosmic and macroscopic is lacking, and most of the existing systems focus on acquiring macroscopic operation parameters such as temperature, pressure, vibration and the like, or independently apply a certain nondestructive detection technology (such as ultrasonic waves and vortex) to carry out offline or online detection. The quantitative association between the macroscopic operation working condition and the material microstructure degradation (such as creep holes and fatigue crack initiation) is difficult to establish by the method, and the deep perception of the damage mechanism cannot be realized. For the heating surface of the boiler, the existing monitoring is often carried out in a scattered way, such as wall thickness measurement, coarse inflation inspection or endoscopy inspection is carried out independently, the data are isolated from each other, and the integral health state of the pipe and the coupling influence among different damage modes are difficult to comprehensively evaluate. For example, the national electric power patent focuses on predicting the risk of scale shedding through steam temperature and thermal load, and the chinese patent optimizes the corrosion monitoring path through a motion sensor, all for specific failure modes, lacks systemicity. The evaluation model is isolated, the mechanism is separated from the data driving, the traditional state evaluation either depends on a mechanism model based on a physical formula (such as Larson-Miller parameter method for evaluating creep life), the accuracy of which is limited by model simplification and parameter uncertainty, or completely depends on a data driving statistical or machine learning model, the interpretation of which is poor, and the extrapolation reliability is low in the absence of a fault sample or in the presence of severe changes of working conditions. Although Hubei Hua electric patent introduces parameter association and dynamic weight between devices, the physical failure mechanism of the material layer is not deeply integrated into the evaluation process. For the heating surface of the boiler, a comprehensive evaluation framework capable of simultaneously integrating a plurality of mechanism models such as wall thickness reduction, expansion and thickness, oxide skin growth, tissue aging and the like and an operation data driving model is not available. The information fusion is insufficient, the decision support has low intelligence level, the monitoring data sources are various (DCS, SIS, spot inspection and nondestructive inspection), the formats and the frequencies are different, and an 'information island' is formed. The prior art lacks an effective multi-source heterogeneous data fusion method, and cannot fully mine complementarity and synergistic value among data. For the heated surface of the boiler, effective space-time alignment and correlation analysis is lacking among the operation parameters (such as wall temperature and smoke temperature), off-line detection data (such as wall thickness, metallographic phase) an