Experience-based decision support methodology for manufacturing technology selection : a fuzzy-decision-tree mining approach

Manufacturing companies must invest in new technologies and processes to succeed in a rapidly changing global environment. Managers have the difficulty of justifying capital investment in adopting new, state-of-the-art technology. Technology investment accounts for a large part of capital spending and is a key form of improving competitive advantage. Typical approaches focus on the expected return of investment and financial reward gained from the implementation of such equipment. With an increasingly dynamic market environment and global economic model, forecasting of financial payback can be argued to become increasingly less accurate. Subsequently, less quantifiable factors are becoming increasingly important. For example, the alignment of a technology with an organisations objective to fulfil future potential and gain competitive advantage is becoming as crucial as economic evaluation. In addition, the impact on human operators and skill level required must be considered. This research was motivated by the lack of decision methodologies that understand why a technology is more successful within an environment rather than re-examining the underlying performance attributes of a technology. The aim is to create a common approach where both experts and non-experts can use historical decision information to support the evaluation and selection of an optimal manufacturing technology. This form of approach is based on the logic in which a decision maker would irrationally recall previous decisions to identify relationships with new problem cases. The work investigates data mining and machine learning techniques to discover the underlying influences to improve technology selection under a set of dynamic factors. The approach initially discovers the practices to which an expert would conduct the selection of a manufacturing technology within industry. A defined understanding of the problem and techniques was subsequently concluded. This led to an understanding of the structure by which historical decision information is recalled by an expert to support new selection problems. The key attributes in the representation of a case were apparent and a form of characterising tangible and intangible variables was justified. This led to the development of a novel, experience-based manufacturing technology selection framework using fuzzy-decision-trees. The methodology is an iterative approach of learning from previously implemented technology cases. Rules and underlying knowledge of the relationships in past cases predicts the outcome of new decision problems. The link of information from a multitude of historical cases may identify those technologies with technical characteristics that perform optimally for projects with unique requirements. This also indicates the likeliness of technologies performing successfully based on the project requirements. Historical decision cases are represented through original project objectives, technical performance attributes of the chosen technology and judged project performance. The framework was shown to provide a comprehensive foundation for decision support that reduces the uncertainty and subjective influence within the selection process. The model was developed with industrial guidance to represent the actions of a manufacturing expert. The performance of the tool was measured by industrial experts. The approach was found to represent well the decision logic of a human expert based on their developed experience through cases. The application to an industrial decision case study demonstrated encouraging results and use by decision makers feasible. The model reduces the subjectivity in the process by using case information that is formed from multiple experts of a prior decision case. The model is applied in a shorter time period than existing practices and the ranking of potential solutions is well aligned to the understanding of a decision maker. To summarise, this research highlights the importance of focusing on less quantifiable factors and the performance of a technology to a specific problem/environment. The arrangement of case information thus represents the experience an expert would acquire and recall as part of the decision process.