Particle Filter with Operational-Scalable Takagi–Sugeno Fuzzy Degradation Model for Filter-Clogging Prognosis

Abstract

In this paper, filter clogging is used as an aerospace integrated vehicle health management case study to demonstrate the proposed prognostic approach. The focus of this paper is on a scalable data-driven degradation model and how it can improve the remaining useful life prediction performance in condition monitoring of a filter component. Instead of overall fitting of the data, a degradation pattern (a parameterized Takagi-Sugeno fuzzy model) is learned from experimental data collected under a range of operating conditions in the proposed approach. The parameter allows the model to scale to fit different degradation profiles, and hence a more accurate model. In realtime condition monitoring, the degradation and model parameter are simultaneously estimated online based on noisy measurement updates using a particle filter. The estimation results show close tracking of the degradation state and good convergence of the model parameter to its real value. The remaining useful life prediction results show low prediction errors, regardless of operating conditions, which contrasts to a conventional data-driven model (a nonparameterized Takagi-Sugeno fuzzy model) where prediction errors increase as operating conditions deviate from the nominal condition. I. IntroductionI NTEGRATED vehicle health management (IVHM) is a major component in a new future aerospace asset management paradigm where a conscious effort is made to shift asset maintenance from a scheduled-based approach to a more proactive and predictive approach. Its goal is to maximize asset operational availability while minimizing downtime and the logistics footprint through monitoring deterioration of component conditions. IVHM involves data processing, which comprehensively consists of capturing data related to assets, monitoring parameters, assessing current or future health conditions through prognostics, and providing recommended maintenance actions.The fuel system is a critical subsystem of the mobile platform asset type, like aircraft, where its functional failures can lead to an aircraft returning to ground or diverting and a remote possibility of engine shutdown. In an economic term, this means a significant cost to an operator. Several IVHM-related examples of fuel systems are reported in the literature: notably, those based on classification and estimation techniques [1], the applied support vector machine, k-nearest neighbors, and the Bayes classifier in fault detection and isolation (FDI) of an experimental electronic control fuel system. In [2,3], a rule-based expert system is developed for FDI of an aircraft fuel system. The Kalman filter is different, as it is an online parameter estimation technique and has been applied to the FDI of an aircraft fuel system [4] and a diesel engine [5]. The previously mentioned literature, however, only addresses FDI problems. It does not predict deterioration of a system, which is key to predictive maintenance.Until now, [6] was the only study related to prognostics of the fuel system; however, the results show s...

Data-driven Approach for RUL Prediction of an Experimental Filtration System

“…We are aware that the modelbased aspect could be improved by considering a more precise physics-based clogging progression model (Eker et al, 2015) and a more accurate fitting of the parameters characterizing the pressure-drop curves, e.g. with particle filtering, as done in (Sreenuch et al, 2015). Such improvements would be relevant especially when particle size and concentration cannot be known exactly and must be inferred from the observed data.…”

“…Since the definition of RUL of the filter is based on the time instant at which the pressure drop reaches the threshold of 20psi, our approach is based on the analysis of the time evolution of the curves representing that drop (Sreenuch et al, 2015). We characterized the drop evolution by means of several statistical features such as kurtosis (kurt), meanto-peak ratio (mpr), root mean square (rms), root mean square in frequency-domain (rmsf), wavelet spectral energy (wse), skewness (skew), variance (var), standard deviation (std), covariance (cov) and the slope of the lin- 2019).…”

“…In our approach, we did not try to model the degradation curves directly, as done for example with a parameterized model in (Sreenuch, Khan, & Li, 2015). We adopted a mainly data-driven approach, where the curves are learned from the experimental data during the training phase, but we also exploited to some extent a model-based component consisting in the computation of the amount of clogging particles progressively intercepted by the filter.…”

Creation of Publicly Available Data Sets for Prognostics and Diagnostics Addressing Data Scenarios Relevant to Industrial Applications

“…For instance, the global industrial air filtration market in 2019 is estimated at USD 3.19 billion (Statista Research Department, 2019). Consequently, filtration has also been studied in PHM; for example, in Eker, Camci, and Jennions (2016), Sreenuch, Khan, and Li (2015), Saarela, Hulsund, Taipale, andHegle (2014), andSkaf, Eker, andJennions (2017). The main focus of these works is placed on the filtration of solids from liquids, more precisely the fuel filtration of vehicles.…”

Modelling the clogging of gas turbine filter houses in heavy-duty power generation systems

“…Apart from the above-mentioned studies [8,9,[12][13][14], several successful modelling attempts on the filtration/clogging process can be found in the literature [10,[15][16][17][18][19][20][21]. Furthermore, Sreenuch et al [22] proposed a scalable data-driven degradation pattern (a parameterized Takagi-Sugeno fuzzy model) to highlight the potential of the prognostic approach in a real-world filter-clogging case study. In real-time condition monitoring, the authors used a filter-clogging experiment as an aerospace application, and the degradation and model parameter were simultaneously forecasted online according to noisy measurement updates using a particle filter.…”