Volker Smits , Christopher Illg , Hermann Klein , Oliver Nelles
{"title":"Analysis of space-filling excitation signals and dynamic models for nonlinear system identification of dynamic processes of a Diesel engine","authors":"Volker Smits , Christopher Illg , Hermann Klein , Oliver Nelles","doi":"10.1016/j.conengprac.2025.106454","DOIUrl":null,"url":null,"abstract":"<div><div>The paper investigates the influence of design of experiments (DoE) methods in combination with various mathematical architectures for the goal of an accurate Diesel engine identification. Three step-based excitation signals – amplitude-modulated pseudo random binary signal (APRBS), optimized nonlinear input signal (OMNIPUS), and the stacked global optimized amplitude time signal (sGOATS) – are analyzed regarding their space-filling property, a good excitation of all frequencies, and achievable model quality. Six different model architectures are used: a finite impulse response multilayer perceptron (MLP-NFIR), a long short-term memory (LSTM) network, a gated recurrent unit (GRU) network, a local model network with regularized finite impulse response models (LMN-NRFIR), a local model network with auto-regressive with exogenous inputs models (LMN-NARX), and a local model state space network (LMSSN). The model architectures are designed to be suitable for low-performance microcontrollers. As a process, a real-world system (Diesel engine) with high system complexity due to multiple feedback paths, strong nonlinearities, and infeasible regions is chosen, which allows a proper investigation of the data-based methods. The sGOATS achieves the best space-filling property, an even excitation of all frequencies, and provides good model qualities. It is shown that it is worth to optimize the excitation signals in order to get better models. Regarding the model architectures, the LMN-NARX and LMN-NRFIR perform worst, whereas the LSTM, GRU, and LMSSN slightly surpass the MLP-NFIR.</div></div>","PeriodicalId":50615,"journal":{"name":"Control Engineering Practice","volume":"164 ","pages":"Article 106454"},"PeriodicalIF":5.4000,"publicationDate":"2025-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Control Engineering Practice","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0967066125002163","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"AUTOMATION & CONTROL SYSTEMS","Score":null,"Total":0}
引用次数: 0
Abstract
The paper investigates the influence of design of experiments (DoE) methods in combination with various mathematical architectures for the goal of an accurate Diesel engine identification. Three step-based excitation signals – amplitude-modulated pseudo random binary signal (APRBS), optimized nonlinear input signal (OMNIPUS), and the stacked global optimized amplitude time signal (sGOATS) – are analyzed regarding their space-filling property, a good excitation of all frequencies, and achievable model quality. Six different model architectures are used: a finite impulse response multilayer perceptron (MLP-NFIR), a long short-term memory (LSTM) network, a gated recurrent unit (GRU) network, a local model network with regularized finite impulse response models (LMN-NRFIR), a local model network with auto-regressive with exogenous inputs models (LMN-NARX), and a local model state space network (LMSSN). The model architectures are designed to be suitable for low-performance microcontrollers. As a process, a real-world system (Diesel engine) with high system complexity due to multiple feedback paths, strong nonlinearities, and infeasible regions is chosen, which allows a proper investigation of the data-based methods. The sGOATS achieves the best space-filling property, an even excitation of all frequencies, and provides good model qualities. It is shown that it is worth to optimize the excitation signals in order to get better models. Regarding the model architectures, the LMN-NARX and LMN-NRFIR perform worst, whereas the LSTM, GRU, and LMSSN slightly surpass the MLP-NFIR.
期刊介绍:
Control Engineering Practice strives to meet the needs of industrial practitioners and industrially related academics and researchers. It publishes papers which illustrate the direct application of control theory and its supporting tools in all possible areas of automation. As a result, the journal only contains papers which can be considered to have made significant contributions to the application of advanced control techniques. It is normally expected that practical results should be included, but where simulation only studies are available, it is necessary to demonstrate that the simulation model is representative of a genuine application. Strictly theoretical papers will find a more appropriate home in Control Engineering Practice''s sister publication, Automatica. It is also expected that papers are innovative with respect to the state of the art and are sufficiently detailed for a reader to be able to duplicate the main results of the paper (supplementary material, including datasets, tables, code and any relevant interactive material can be made available and downloaded from the website). The benefits of the presented methods must be made very clear and the new techniques must be compared and contrasted with results obtained using existing methods. Moreover, a thorough analysis of failures that may happen in the design process and implementation can also be part of the paper.
The scope of Control Engineering Practice matches the activities of IFAC.
Papers demonstrating the contribution of automation and control in improving the performance, quality, productivity, sustainability, resource and energy efficiency, and the manageability of systems and processes for the benefit of mankind and are relevant to industrial practitioners are most welcome.