{"title":"利用模拟传感器记录的训练数据集形成方法确定直升机涡轴发动机效率","authors":"Denys Baranovskyi , Serhii Vladov , Maryna Bulakh , Valerii Sokurenko , Oleksandr Muzychuk , Victoria Vysotska","doi":"10.1016/j.ymssp.2025.113368","DOIUrl":null,"url":null,"abstract":"<div><div>This article presents a novel approach to intelligent monitoring and control of complex dynamic systems, focusing specifically on helicopter turboshaft engines during flight. The developed method includes multi-channel signal processing, adaptive discretization and quantization, temporal feature extraction, and singular spectrum analysis. The key components are median, mean, and Hilbert filtering to eliminate noise, as well as adaptive quantization and clustering (hierarchical, DBSCAN, Gaussian mixtures) to ensure homogeneity and representativeness of samples. The method was verified using a neural network model, demonstrating a mean square error (MSE) of no more than 0.025 on training, validation, and test data. As a numerical experiment part, the TV3-117 engine compressor’s efficiency installed on the Mi-8MTV helicopter was calculated. The results showed a maximum MSE deviation from the reference value of no more than 0.862%, which confirms the developed method’s high accuracy. The article also proves the theorem on homogeneity and representativeness of data, according to which, if the training and test samples satisfy the homogeneity criteria (according to the Fisher-Pearson and Fisher-Snedecor statistical criteria) and representativeness (according to cluster analysis), they can be considered suitable for use in practical problems of modeling, classification, and forecasting. This theorem’s theoretical justification confirms the need for strict quality control of samples before training models. Scenarios with artificial introduction of errors into the data were simulated, which led to calculations and confirmed the instability and importance of ensuring homogeneity and representativeness of datasets. The developed method allows for a significant increase in the accuracy of predicting and diagnosing anomalies in the helicopter turboshaft engine’s operation, providing a reliable basis for intelligent monitoring and control in real operating conditions.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"239 ","pages":"Article 113368"},"PeriodicalIF":8.9000,"publicationDate":"2025-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"The method for training dataset forming recorded by analog sensors to determine the helicopter turboshaft engines efficiency\",\"authors\":\"Denys Baranovskyi , Serhii Vladov , Maryna Bulakh , Valerii Sokurenko , Oleksandr Muzychuk , Victoria Vysotska\",\"doi\":\"10.1016/j.ymssp.2025.113368\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>This article presents a novel approach to intelligent monitoring and control of complex dynamic systems, focusing specifically on helicopter turboshaft engines during flight. The developed method includes multi-channel signal processing, adaptive discretization and quantization, temporal feature extraction, and singular spectrum analysis. The key components are median, mean, and Hilbert filtering to eliminate noise, as well as adaptive quantization and clustering (hierarchical, DBSCAN, Gaussian mixtures) to ensure homogeneity and representativeness of samples. The method was verified using a neural network model, demonstrating a mean square error (MSE) of no more than 0.025 on training, validation, and test data. As a numerical experiment part, the TV3-117 engine compressor’s efficiency installed on the Mi-8MTV helicopter was calculated. The results showed a maximum MSE deviation from the reference value of no more than 0.862%, which confirms the developed method’s high accuracy. The article also proves the theorem on homogeneity and representativeness of data, according to which, if the training and test samples satisfy the homogeneity criteria (according to the Fisher-Pearson and Fisher-Snedecor statistical criteria) and representativeness (according to cluster analysis), they can be considered suitable for use in practical problems of modeling, classification, and forecasting. This theorem’s theoretical justification confirms the need for strict quality control of samples before training models. Scenarios with artificial introduction of errors into the data were simulated, which led to calculations and confirmed the instability and importance of ensuring homogeneity and representativeness of datasets. The developed method allows for a significant increase in the accuracy of predicting and diagnosing anomalies in the helicopter turboshaft engine’s operation, providing a reliable basis for intelligent monitoring and control in real operating conditions.</div></div>\",\"PeriodicalId\":51124,\"journal\":{\"name\":\"Mechanical Systems and Signal Processing\",\"volume\":\"239 \",\"pages\":\"Article 113368\"},\"PeriodicalIF\":8.9000,\"publicationDate\":\"2025-09-17\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Mechanical Systems and Signal Processing\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0888327025010696\",\"RegionNum\":1,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"ENGINEERING, MECHANICAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Mechanical Systems and Signal Processing","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0888327025010696","RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, MECHANICAL","Score":null,"Total":0}
The method for training dataset forming recorded by analog sensors to determine the helicopter turboshaft engines efficiency
This article presents a novel approach to intelligent monitoring and control of complex dynamic systems, focusing specifically on helicopter turboshaft engines during flight. The developed method includes multi-channel signal processing, adaptive discretization and quantization, temporal feature extraction, and singular spectrum analysis. The key components are median, mean, and Hilbert filtering to eliminate noise, as well as adaptive quantization and clustering (hierarchical, DBSCAN, Gaussian mixtures) to ensure homogeneity and representativeness of samples. The method was verified using a neural network model, demonstrating a mean square error (MSE) of no more than 0.025 on training, validation, and test data. As a numerical experiment part, the TV3-117 engine compressor’s efficiency installed on the Mi-8MTV helicopter was calculated. The results showed a maximum MSE deviation from the reference value of no more than 0.862%, which confirms the developed method’s high accuracy. The article also proves the theorem on homogeneity and representativeness of data, according to which, if the training and test samples satisfy the homogeneity criteria (according to the Fisher-Pearson and Fisher-Snedecor statistical criteria) and representativeness (according to cluster analysis), they can be considered suitable for use in practical problems of modeling, classification, and forecasting. This theorem’s theoretical justification confirms the need for strict quality control of samples before training models. Scenarios with artificial introduction of errors into the data were simulated, which led to calculations and confirmed the instability and importance of ensuring homogeneity and representativeness of datasets. The developed method allows for a significant increase in the accuracy of predicting and diagnosing anomalies in the helicopter turboshaft engine’s operation, providing a reliable basis for intelligent monitoring and control in real operating conditions.
期刊介绍:
Journal Name: Mechanical Systems and Signal Processing (MSSP)
Interdisciplinary Focus:
Mechanical, Aerospace, and Civil Engineering
Purpose:Reporting scientific advancements of the highest quality
Arising from new techniques in sensing, instrumentation, signal processing, modelling, and control of dynamic systems