A framework for improving process robustness with quantification of uncertainties in Industry 4.0

Digitalisation of industrial processes, also called the fourth industrial revolution, is leading to availability of large volume of data containing measurements of many process variables. This offers new opportunities to gain deeper insights on process variability and its effects on quality and performance. Manufacturing facilities already use data driven approaches to study process variability and find improvement opportunities through methodologies such as Design of Experiment (DOE) and Six Sigma. However, current approaches are not adequate to model the complexity of modern manufacturing systems, especially when these systems exhibit non-linear interactions between high numbers of variables. In this paper a methodology to improve process robustness is proposed. This methodology uses non-parametric estimation of quantiles of response to discover new tolerance limits of factors. This method does not make any stringent assumption of linearity and works well in finding the interactions effects of covariates on response quantiles. Process robustness, which is defined as the ability of a process to have acceptable quality whilst tolerating variability of the input, is measured through calculation of Likelihood Ratios (LR) associated to the new tolerance limits. Uncertainty of this estimation is quantified via simulations using the bootstrapping method. The novel contribution of this paper is the application of quantile regression and likelihood ratios to the tolerance synthesis problem applied to a low alloy foundry. It shows the validity of the methodology in modelling behaviours of complex manufacturing processes using data driven approaches to gain new insights on causes of process variabilities and discover new product specific process knowledge. This work contributes to bridging the gap between theory and application towards implementing Industry 4.0 predictive analytics.

[1]  Robert Tibshirani,et al.  Bootstrap Methods for Standard Errors, Confidence Intervals, and Other Measures of Statistical Accuracy , 1986 .

[2]  Meghana R. Ransing,et al.  A quality correlation algorithm for tolerance synthesis in manufacturing operations , 2016, Comput. Ind. Eng..

[3]  Inci Batmaz,et al.  A review of data mining applications for quality improvement in manufacturing industry , 2011, Expert Syst. Appl..

[4]  Mohsen Ostad Shabani,et al.  The ANN application in FEM modeling of mechanical properties of Al–Si alloy , 2011 .

[5]  Xiang Chen,et al.  Numerical optimization of gating system parameters for a magnesium alloy casting with multiple performance characteristics , 2008 .

[6]  Nur Evin Özdemirel,et al.  Defect Cause Modeling with Decision Tree and Regression Analysis , 2007 .

[7]  Rajesh Ransing,et al.  Knowledge management and knowledge discovery for process improvement and sustainable manufacturing: a foundry case study , 2016 .

[8]  Goran Radenković,et al.  Application of ANOVA method to precipitation behaviour studies , 2005 .

[9]  Michael Glodek,et al.  Process Robustness - A PQRI White Paper , 2006 .

[10]  Rajesh Ransing.  “If Only My Foundry Knew What it Knows ...”: A 7Epsilon Perspective on Root Cause Analysis and Corrective Action Plans for ISO9001:2008 , 2014 .

[11]  Luis E. Zárate,et al.  Qualitative behavior rules for the cold rolling process extracted from trained ANN via the FCANN method , 2009, Eng. Appl. Artif. Intell..

[12]  Liangxiao Jiang,et al.  An Empirical Study on Class Probability Estimates in Decision Tree Learning , 2011, J. Softw..

[13]  N. Jawahar,et al.  Process factor optimization for controlling pull-down defects in iron castings , 2009 .

[14]  Ivan Bratko,et al.  On Estimating Probabilities in Tree Pruning , 1991, EWSL.

[15]  Meghana R. Ransing,et al.  Organisational Knowledge Management for Defect Reduction and Sustainable Development in Foundries , 2015, Int. J. Knowl. Syst. Sci..

[16]  R. Koenker,et al.  Regression Quantiles , 2007 .

[17]  Meghana R. Ransing,et al.  A novel variable selection approach based on co-linearity index to discover optimal process settings by analysing mixed data , 2014, Comput. Ind. Eng..

[18]  Dirk Cattrysse,et al.  Cost estimation for sheet metal parts using multiple regression and artificial neural networks: A case study , 2008 .

[19]  Gisela Lanza,et al.  The Concept of Robustness in Production Systems and its Correlation to Disturbances , 2014 .

[20]  J A Hanley,et al.  Bootstrap confidence intervals for the sensitivity of a quantitative diagnostic test. , 2000, Statistics in medicine.

[21]  Shubhabrata Datta,et al.  Designing cold rolled IF steel sheets with optimized tensile properties using ANN and GA , 2011 .

[22]  A. Bahrami,et al.  Prediction of porosity percent in Al–Si casting alloys using ANN , 2006 .

[23]  Meghana R. Ransing,et al.  A coupled penalty matrix approach and principal component based co-linearity index technique to discover product specific foundry process knowledge from in-process data in order to reduce defects , 2013, Comput. Ind..

[24]  Shahram Abbasi,et al.  Artificial Neural Network Modeling the Tensile Strength of Hot Strip Mill Products , 2009 .

[25]  Rajesh S. Ransing,et al.  Risk based uncertainty quantification to improve robustness of manufacturing operations , 2016, Comput. Ind. Eng..