Variable importance assessment in lifespan models of insulation materials: A comparative study

Abstract

This paper presents and compares different methods for evaluating the relative importance of variables involved in insulation lifespan models. Parametric and nonparametric models are derived from accelerated aging tests on twisted pairs covered with an insulating varnish under different stress constraints (voltage, frequency and temperature). Parametric models establish a simple stresslifespan relationship and the variable importance can be evaluated from the estimated parameters. As an alternative approach, non-parametric models explain the stress-lifespan relationship by means of regression trees or random forests (RF) for instance. Regression trees naturally provide a hierarchy between the variables. However, they suffer from a high dependency with respect to the training set. This paper shows that RF provide a more robust model while allowing a quantitative variable importance assessment. Comparisons of the different models are performed on different training and test sets obtained through experiments.

Parametric and Nonparametric Models for Lifespan Modeling of Insulation Systems in Electrical Machines

“…The most efficient method to evaluate the effects of several factors and their interactions on a response variable is the basic DoE [10] that has been already validated for insulation lifespan models in [9], [11], [12], [13]. According to DoE [9], [10], 2 levels ( 1) are assigned to each factor.…”

“…Of course, the type of insulation material will affect the model coefficients but not the lifespan modeling ability of the selected methods. Indeed, it has been proven with different materials and different levels of the stress factors in [9], [11], [12], [13] that the methods based either on DoE or on RS lead to satisfying lifespan modeling results. This section will investigate the effect of the initial conditions of the insulation material on the results.…”

“…The objective is to provide a reliable lifespan model with a minimum experimental cost for economic purposes. To comply with both the economical and the accuracy constraints, the number of experiments and their configuration were specified in previous work according to two experimental optimization methods: the Design of Experiments (DoE) [9]- [13] and the Response Surface (RS) [9], [11], [12], [14]. The proposed insulation lifespan models included three different stress factors: voltage, frequency and temperature, as they were identified as the predominant factors causing PD [3], [4].…”

The Impact of Thermal Degradation on Properties of Electrical Machine Winding Insulation Material

“…Also, in [25] a new cable monitoring method was proposed based on impedance spectrum analysis in the high frequency range. Furthermore, in [26] the authors used a regression tree constructed with 32 experiments for insulation lifespan modelling (twisted pairs). One of their findings is that at low voltages, only the temperature has a significant effect on the lifespan.…”

“…If the applied temperature exceeds some critical value then different chemical reactions are caused compared to the real S ageing mechanisms. However, in [26] the application of multi-stress ageing on stator bars was successful in predicting the real ageing mechanisms of actual bars in applications working for 22 years. In the same work, the significant role of mechanical stress is highlighted because the loosening of the stator bars in the slots leads to vertical vibrations which enhance the degradation mechanism.…”