Methods for equipment health index estimation and prediction in case of imbalanced samples of heterogeneous data

  • 基本信息
其他名称: 24-21-00483
项目负责人: Shcherbakov Maxim
发表日期: 2024
主持机构: Federal state budget educational of office highly professional education "Volgograd State Technical University",
国家: 俄罗斯
开始日期: 2024
结束日期: 2025
简介: The project is aimed at solving the problem of creating and using effective information technologies with artificial intelligence components to ensure the reliability and fault tolerance of technical systems.$The relevance of solving the problem of creating an effective control of the technical condition of equipment can be justified by the presence of the following limitations.$1. Existing formal methods for determining the technical condition are requirements of the regulatory authority rather than the operator and cannot characterize the actual technical condition in terms of ensuring reliability. 2. Training of models for assessing and forecasting the technical condition is usually carried out on a sample of data obtained from other similar objects. However, the question arises as to the applicability of customized models to other objects.3. Existing approaches aimed at predicting the remaining resource based on aggregated data of work cycles and methods for obtaining such aggregated data require significant participation of experts. 4. The cyclicity and different duration of the operating cycle of equipment operation lead to restrictions on the use of forecasting models that analyze fixed segments of time series (and such models are a majority). 5. In practice, the data are high-frequency multivariate time series, the number variables in which corresponds to the number of process parameters. Using the most advanced state-of-the-art models (eg, convolutional neural networks and hybrid architectures) does not lead to satisfactory results. 6. The heterogeneity of data sources and the data itself, characterized as weakly structured or unstructured, imposes significant restrictions on the use of approaches for assessing and predicting the technical condition. In addition, the presence of both qualitative and quantitative (often when describing operating conditions) characteristics makes it difficult to divide the data sample into training and test sets. 7 The imbalance of the data sample is characterized by the fact that the number of observations with failures is much smaller than the number of observations in a normal working state. In this regard, the solution of the problem by reducing it to the problem of regression analysis, even by highly efficient approaches, leads to unsatisfactory results. We propose the following sentence which aggregates mentioned above limitations: in case of imbalanced samples of heterogeneous data. $Scientific contribution is new methods for assessing and predicting the health index of equipment aimed at effective reliability management, including models and algorithms for processing unbalanced samples of heterogeneous data. In particular, it is planned to obtain the following world-class scientific results: a model of the health index of complex equipment; a preprocessing method of high-frequency technological parameters data; criterion for assessing the similarity of objects for the task health index prediction for objects of the same type; training and test (cross-validation) samples design method; method for identifying the degradation process in the operation of equipment; health index prediction method; remaining useful life forecasting method; a software framework that implements the proposed approaches.$The solution of the problem presented in the project will allow: 1. increase the efficiency of designing a new class of equipment as a cyber-physical system, for which technical solutions for collecting data characterizing the technical condition, methods for assessing and predicting the technical condition to ensure trouble-free operation are provided; 2. extend the life of existing equipment by developing optimal maintenance and repair strategies focused on ensuring reliability; 3. to extend the resource of existing equipment by forming an optimal mode of operation for a) reducing the probability of failures (or limit state) and b) extending the resource (increasing the time until the limit state occurs).
专业领域: 信息技术
语种: 英语

相关项目

Methods for equipment health index estimation and prediction in case of imbalanced samples of heterogeneous data

The project is ... 2024