Maximal overlap discrete wavelet packet transforms and variants of neutrosophic cubic cross-entropy-based identification of rotor defects

Abstract

The present study proposes a novel procedure for automated testimony of rotor defects through maximal overlap discrete wavelet packet transform (MODWPT) and the proposed neutrosophic cubic cross-entropy, fuzzy cross-entropy and single-valued neutrosophic cross-entropy (NCE) measures consecutively. Discrete wavelet transform is an efficient data decomposition technique, but the technical barrier with this technique is that it can decompose only low-frequency (approximate) signals and also it does not possess the shift invariance property. MODWPT possesses the shift invariance property and is highly capable of decomposing both approximate and detailed signals in time–frequency analysis. After the decomposition of raw signals through MODWPT, the lower and upper bounds of normalized energy readings of various rotor defect conditions (familiar and unfamiliar) are extracted and converted into the forms of fuzzy sets, single-valued neutrosophic sets and neutrosophic cubic sets respectively. The minimum neutrosophic cubic cross-entropy measure as well as fuzzy and single-valued NCE measure values between each rotor defect condition (familiar and unfamiliar) are computed and then utilized to identify various rotor defect conditions, such as parallel and angular misalignment of 7 Mils, rub, unbalance and defect-free respectively. The proposed variants of cross-entropy measure are intelligent in identifying the rotor defect conditions in comparison with the existing cosine similarity measure which exhibits unjustified and meaningless results during mathematical treatments.