An early-stage decision-support framework for the implementation of intelligent automation

The constant pressure on manufacturing companies to improve productivity, reduce the lead time and progress in quality requires new technological developments and adoption.

The rapid development of smart technology and robotics and autonomous systems (RAS) technology has a profound impact on manufacturing automation and might determine winners and losers of the next generation’s manufacturing competition. Simultaneously, recent smart technology developments in the areas enable an automation response to new production paradigms such as mass customisation and product-lifecycle considerations in the context of Industry 4.0. New paradigms, like mass customisation, increased both the complexity of the tasks and the risk due to smart technology integration. From a manufacturing automation perspective, intelligent automation has been identified as a possible response to arising demands. The presented research aims to support the industrial uptake of intelligent automation into manufacturing businesses by quantifying risks at the early design stage and business case development. An early-stage decision-support framework for the implementation of intelligent automation in manufacturing businesses is presented in this thesis.

The framework is informed by an extensive literature review, updated and verified with surveys and workshops to add to the knowledge base due to the rapid development of the associated technologies. A paradigm shift from cost to a risk-modelling perspective is proposed to provide a more flexible and generic approach applicable throughout the current technology landscape. The proposed probabilistic decision-support framework consists of three parts:

• A clustering algorithm to identify the manufacturing functions in manual processes from task analysis to mitigate early-stage design uncertainties

• A Bayesian Belief Network (BBN) informed by an expert elicitation via the DELPHI method, where the identified functions become the unit of analysis.

• A Markov-Chain Monte-Carlo method modelling the effects of uncertainties on the critical success factors to address issues of factor interdependencies after expert elicitation.

Based on the overall decision framework a toolbox was developed in Microsoft Excel. Five different case studies are used to test and validate the framework. Evaluation of the results derived from the toolbox from the industrial feedback suggests a positive validation for commercial use. The main contributions to knowledge in the presented thesis arise from the following four points:

• Early-stage decision-support framework for business case evaluation of intelligent automation.

• Translating manual tasks to automation function via a novel clustering approach

• Application of a Markov-Chain Monte-Carlo Method to simulate correlation between decision criteria

• Causal relationship among Critical Success Factors has been established from business and technical perspectives.

The implications on practise might be promising. The feedback arising from the created tool was promising from the industry, and a practical realisation of the decision-support tool seems to be desired from an industrial point of view.

With respect to further work, the decision-support tool might have established a ground to analyse a human task automatically for automation purposes. The established clustering mechanisms and the related attributes could be connected to sensorial data and analyse a manufacturing task autonomously without the subjective input of task analysis experts. To enable such an autonomous process, however, the psychophysiological understanding must be increased in the future.