Search

Manufacturing is currently experiencing a paradigm shift in the way that products are designed, produced and serviced. Such changes are brought about mainly by the extensive use of the Internet and digital technologies. As a result of this shift, a new industrial revolution is emerging, termed “Industry 4.0” (i4), which promises to accommodate mass customisation at a mass production cost. For i4 to become a reality, however, multiple challenges need to be addressed, highlighting the need for design for agile manufacturing and, for this, a framework capable of integrating big data analytics arising from the service end, business informatics through the manufacturing process, and artificial intelligence (AI) for the entire manufacturing value chain. This thesis attempts to address these issues, with a focus on the need for design for agile manufacturing. First, the state of the art in this field of research is reviewed on combining cutting-edge technologies in digital manufacturing with big data analysed to support agile manufacturing. Then, the work is focused on developing an AI-based framework to address one of the customisation issues in smart design and agile manufacturing, that is, prediction of potential customer needs and wants. With this framework, an AI-based approach is developed to predict design attributes that would help manufacturers to decide the best virtual designs to meet emerging customer needs and wants predictively. In particular, various machine learning approaches are developed to help explain at least 85% of the design variance when building a model to predict potential customer needs and wants. These approaches include k-means clustering, self-organizing maps, fuzzy k-means clustering, and decision trees, all supporting a vector machine to evaluate and extract conscious and subconscious customer needs and wants. A model capable of accurately predicting customer needs and wants for at least 85% of classified design attributes is thus obtained. Further, an analysis capable of determining the best design attributes and features for predicting customer needs and wants is also achieved. As the information analysed can be utilized to advise the selection of desired attributes, it is fed back in a closed-loop of the manufacturing value chain: design → manufacture → management/service → → → design ... For this, a total of 4 case studies are undertaken to test and demonstrate the efficacy and effectiveness of the framework developed. These case studies include: 1) an evaluation model of consumer cars with multiple attributes including categorical and numerical ones; 2) specifications of automotive vehicles in terms of various characteristics including categorical and numerical instances; 3) fuel consumptions of various car models and makes, taking into account a desire for low fuel costs and low CO2 emissions; and 4) computer parts design for recommending the best design attributes when buying a computer. The results show that the decision trees, as a machine learning approach, work best in predicting customer needs and wants for smart design. With the tested framework and methodology, this thesis overall presents a holistic attempt to addressing the missing gap between manufacture and customisation, that is meeting customer needs and wants. Effective ways of achieving customization for i4 and smart manufacturing are identified. This is achieved through predicting potential customer needs and wants and applying the prediction at the product design stage for agile manufacturing to meet individual requirements at a mass production cost. Such agility is one key element in realising Industry 4.0. At the end, this thesis contributes to improving the process of analysing the data to predict potential customer needs and wants to be used as inputs to customizing product designs agilely.