Meghana R. Ransing

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

Foundry process is a complex process with more than 100 parameters that influence the quality of final cast component. It is a process with multiple optimal conditions. For two foundries manufacturing the same alloy and cast geometry, the process and alloy conditions used by one foundry will most likely be different from the other one. For a foundry process engineer, it is also currently difficult to link process knowledge available in the published literature to specific process conditions and defects in a foundry. A concept of product and foundry specific process knowledge has been introduced so that the intellectual property that is created every time a cast component is poured can be stored and reused in order to be able to reduce defects. A methodology has been proposed for discovering noise free correlations and interactions in the data collected during a stable casting process so that small adjustments can be made to several process factors in order to progress towards the zero defects manufacturing environment. The concepts have been demonstrated on actual but anonymised in-process data set on chemical composition for a nickel based alloy.

A novel variable selection approach based on co-linearity index to discover optimal process settings by analysing mixed data

Abstract In the last two decades the application of statistical techniques to process control has gained popularity due to the widespread adoption of quality management systems such as ISO9001. Demonstration of continual process improvement by monitoring process effectiveness has become an integral part of satisfying the requirements of clause 8 of the ISO9001:2008 standard. The process effectiveness is measured in terms of one or more process responses. Data driven approaches are often used to associate the variability in process responses with one or more process variables. However, traditional techniques become unpractical in the presence of large number of variables and noisy data sets. This paper extends the co-linearity index and penalty matrix approach (Ransing et al., 2013) for discovering noise free correlations between heterogeneous process variables and responses. Noise is removed by reducing the dimensionality of the variable space and using robust data pre-treatment methods which are more suitable in the presence of outliers and skewed distributions for process variables. Scaling factors have been proposed to balance variance contributions from response variables, quantitative and categorical variables. The proposed method allows process variables with skewed distribution to contribute more to the variance than Gaussian distributed variables so that these variables can be investigated further, if necessary. Correlations are visualised in a single plot and can be used in real industrial settings to assist process engineers in manufacturing diagnosis and root cause analysis. The applicability and validity of this novel method has been demonstrated through two industrial case studies.

A quality correlation algorithm for tolerance synthesis in manufacturing operations

Abstract The clause 6.1 of the ISO9001:2015 quality standard requires organisations to take specific actions to determine and address risks and opportunities in order to minimize undesired effects in the process and achieve process improvement. This paper proposes a new quality correlation algorithm to optimise tolerance limits of process variables across multiple processes. The algorithm uses reduced p -dimensional principal component scores to determine optimal tolerance limits and also embeds ISO9001:2015s risk based thinking approach. The corresponding factor and response variable pairs are chosen by analysing the mixed data set formulation proposed by Giannetti et al. (2014) and co-linearity index algorithm proposed by Ransing, Giannetti, Ransing, and James (2013). The goal of this tolerance limit optimisation problem is to make several small changes to the process in order to reduce undesired process variation. The optimal and avoid ranges of multiple process parameters are determined by analysing in-process data on categorical as well as continuous variables and process responses being transformed using the risk based thinking approach. The proposed approach has been illustrated by analysing in-process chemistry data for a nickel based alloy for manufacturing cast components for an aerospace foundry. It is also demonstrated how the approach embeds the risk based thinking into the in-process quality improvement process as required by the ISO9001:2015 standard.

Organisational Knowledge Management for Defect Reduction and Sustainable Development in Foundries

Despite many advances in the field of casting technologies the foundry industry still incurs significant losses due to the cost of scrap and rework with adverse effects on profitability and the environment. Approaches such as Six Sigma, DoE, FMEA are used by foundries to address quality issues. However these approaches lack support to manage the heterogeneous knowledge created during process improvement activities. The proposed revision of ISO9001:2015 quality standard puts emphasis on retaining organisational knowledge and its continual use in process improvement. In this paper a novel framework for creation, storage and reuse of product specific process knowledge is presented. The framework is reviewed taking into consideration theoretical perspectives of organisational knowledge management as well as addressing the challenges concerning its practical implementation. A knowledge repository concept is introduced to demonstrate how organisational knowledge can be effectively stored and reused for achieving continual process improvement and sustainable development.

A bootstrap method for uncertainty estimation in quality correlation algorithm for risk based tolerance synthesis

Abstract A risk based tolerance synthesis approach is based on ISO9001:2015 quality standard’s risk based thinking. It analyses in-process data to discover correlations among regions of input data scatter and desired or undesired process outputs. Recently, Ransing, Batbooti, Giannetti, and Ransing (2016) proposed a quality correlation algorithm (QCA) for risk based tolerance synthesis. The quality correlation algorithm is based on the principal component analysis (PCA) and a co-linearity index concept (Ransing, Giannetti, Ransing, & James, 2013). The uncertainty in QCA results on mixed data sets is quantified and analysed in this paper. The uncertainty is quantified using a bootstrap sampling method with bias-corrected and accelerated confidence intervals. The co-linearity indices use the length and cosine angles of loading vectors in a p-dimensional space. The uncertainty for all p-loading vectors is shown in a single co-linearity index plot and is used to quantify the uncertainty in predicting optimal tolerance limits. The effects of re-sampling distributions are analysed. The QCA tolerance limits are revised after estimating the uncertainty in limits via bootstrap sampling. The proposed approach has been demonstrated by analysing in-process data from a previously published case study.

Improving the Gradient Based Search Direction to Enhance Training Efficiency of Back Propagation Based Neural Network Algorithms

Most of the gradient based optimisation algorithms employed during training process of back propagation networks use negative gradient of error as a gradient based search direction. A novel approach is presented in this paper for improving the training efficiency of back propagation neural network algorithms by adaptively modifying this gradient based search direction. The proposed algorithm uses the value of gain parameter in the activation function to modify the gradient based search direction. It has been shown that this modification can significantly enhance the computational efficiency of training process. The proposed algorithm is generic and can be implemented in almost all gradient based optimisation processes. The robustness of the proposed algorithm is shown by comparing convergence rates for gradient descent, conjugate gradient and quasi- Newton methods on many benchmark examples.