Application of Mahalanobis-Taguchi system and design of experiments to reduce the field failures of splined shafts

Abstract

Purpose – The purpose of this paper is to develop a methodology to reduce the field failures of splined shafts. The paper also demonstrates the application of Mahalanobis-Taguchi system (MTS) for identifying the optimum hardness profile to avoid failures. Design/methodology/approach – Through the usage profile analysis and comparison between the failed and good shafts, the major reason for shaft failure was identified as hardness variation. Then MTS approach was used to identify the optimum hardness profile for the shafts. An experiment was designed with power, feed and the gap between inductor and quench ring representing the heat transfer rate, heat removal rate and the time between heat transfer and removal of induction hardening process as factors. Based on experimental results, the optimum combination factors that would reduce the variation around the optimum hardness profile were identified. Findings – The study showed that the shaft failures can be reduced by optimizing the hardness profile of the shafts rather than warning customers on overloading, changing the raw material or investing on machining operation to achieve better shaft finish. The study suggested heat transfer rate, heat removal rate and the time between heat transfer and removal had significant impact on the shaft's hardness profile. The study resulted in reducing the field failures from 0.32 to 0.029 percent. Practical implications – This study provides valuable information on how to identify optimum hardness profile using MTS methodology to reduce shaft failures and how to minimize the variation around the optimum hardness profile using design of experiments. Originality/value – To the best of author's knowledge, no study has been conducted to identify optimum hardness profile using MTS methodology. The study also provides an approach to minimize the variation around a non-linear hardness profile using design of experiments.

Dynamic coal blending under sustainability concerns: a case study from the Indian steel industry

“…Several works consider the application of the Taguchi method in DOE. John (2014) showed shaft failure analysis using Taguchi Method and DOE. Moralı et al (2018) apply the Taguchi method of parameter design with high-specific surface area in the production of activated carbon to optimize the activation process.…”

MD‐based approaches for system health monitoring: a review

“…For components (such as analogue circuits, PCBs, and IGBTs), Vasan et al [42] did an investigation into analogue circuits, developed an MD-based fault indicator in (30), and used a particle filter algorithm for failure prognosis…”

“…Early failures of cooling fan are detected by comparing the HI with MS induction motors fault diagnosis [28] MD and time-frequency classification are used to detect different faults and estimate fault severity by analysing induction motor current signal classify different faults [27] MD is used to indicate induction motor health condition, and Box-Cox transformation and properties of normal distribution are employed to determine MD ranges for different cases centrifugal pump fault detection, isolation, and prognosis [29] MD is used to fuse multiple sensors' data into a single metric for indicating the fault progression. Different faults are then clustered by using the constructed MD metric splined shafts reduce field failures [30] evaluate shafts hardness variation and identify the optimum profile to avoid failures by using MTS gearbox health monitoring and fault detection [31] features are extracted from AE signals using an EMD-based method and are then used for fault detection based on MD statistical method prognosis [32] MD is calculated by using SCADA data to quantify gearbox conditions and fault prognosis fault simulator include motor, shaft and disks intelligent fault diagnosis system for classifying different machine conditions [33] use two-dimensional discrete wavelet transform to extract features from machine's infrared thermal image, then select out features with aid of MD and relief algorithm to improve the classification accuracy aero-engine health monitoring and anomaly detection [34] MD and Weibull distribution are used for engines health monitoring [13] parameters, which are identified by FMMEA and mRMR feature selection algorithm, are used to develop the online MD-based HDD fault detection system failure prediction [35] a two-step parametric approach is proposed to predict HDD failures in which MD is used to fuse multiple variables into one index for indicating HDD's health conditions LED anomaly detection [36] an approach based on MD and Weibull distribution is proposed to detected chromaticity shift of LED real-time health monitoring and prognostics [37] LED's temperature and current data are used to develop an MD-based HI for helping detect anomalies and predict the RUL analogue circuits fault diagnosis [38] an ensemble approach, which integrates cluster technique and MD, is proposed for fault diagnosis in analogue circuit [39] an MD-based approach, which uses the characteristics of analogue circuit, is introduced to analogue circuit fault detection fault detection [40] a test algorithm using the MD-based metric is proposed to detect parametric and catastrophic faults in circuits AC arc fault detection [41] an MD-based approach is proposed to detect AC arc faults fault diagnostics and prognostics [42] analogue circuits fault diagnosis is solved by using SVM, an MDb...…”

An integrated approach for multivariate statistical process control using Mahalanobis-Taguchi System and Andrews function

“…The proposed approach uses MTS, which is basically a multidimensional pattern recognition system, and can be used to distinguish between the normal and abnormal groups (John, 2014). Collecting a set of initial (primary or secondary) in-control and out-of-control data on the identified y (CTQ) variables is a prerequisite to apply MTS.…”