Prediction and improvement of iron casting quality through analytics and Six Sigma approach

Purpose: The purpose of this technical paper is to explore the application of analytics and Six Sigma in the manufacturing processes for iron foundries. This study aims to establish a causal relationship between chemical composition and the quality of the iron casting to achieve the global benchmark quality level. Design/methodology/approach: The case study-based exploratory research design is used in this study. The problem discovery is done through the literature survey and Delphi method-based expert opinions. The prediction model is built and deployed in 11 cases to validate the research hypothesis. The analytics helps in achieving the statistically significant business goals. The design includes Six Sigma DMAIC (Define – Measure – Analyze – Improve and Control) approach, benchmarking, historical data analysis, literature survey and experiments for the data collection. The data analysis is done through stratification and process capability analysis. The logistic regression-based analytics helps in prediction model building and simulations. Findings: The application of prediction model helped in quick root cause analysis and reduction of rejection by over 99 per cent saving over INR6.6m per year. This has also enhanced the reliability of the production line and supply chain with on-time delivery of 99.78 per cent, which earlier was 80 per cent. The analytics with Six Sigma DMAIC approach can quickly and easily be applied in manufacturing domain as well. Research limitations implications: The limitation of the present analytics model is that it provides the point estimates. The model can further be enhanced incorporating range estimates through Monte Carlo simulation. Practical implications: The increasing use of prediction model in the near future is likely to enhance predictability and efficiencies of the various manufacturing process with sensors and Internet of Things. Originality/value: The researchers have used design of experiments, artificial neural network and the technical simulations to optimise either chemical composition or mould properties or melt shop parameters. However, this work is based on comprehensive historical data-based analytics. It considers multiple human and temporal factors, sand and mould properties and melt shop parameters along with their relative weight, which is unique. The prediction model is useful to the practitioners for parameter simulation and quality enhancements. The researchers can use similar analytics models with structured Six Sigma DMAIC approach in other manufacturing processes for the simulation and optimisations.