Knowledge Discovery and Analysis in Manufacturing

ABSTRACT

The quality and reliability requirements for next-generation manufacturing are reviewed, and current approaches are cited. The potential for augmenting current quality/reliability technology is described, and characteristics of potential future directions are postulated. Methods based on knowledge discovery and analysis in manufacturing (KDAM) are reviewed, and related successful applications in business and social fields are discussed. Typical KDAM applications are noted, along with general functions and specific KDAM-related technologies. A systematic knowledge discovery process model is reviewed, and examples of current work are given, including description of successful applications of KDAM to creation of rules for optimizing gas porosity in sand casting molds. Finally, directions in KDAM technology and associated research requirements are described, and comments related to application and acceptance of KDAM are provided.

KEYWORDS:

• data mining
• knowledge discovery in data
• machine learning

Notes

¹When considering technology s-curves, it is essential to note that it is not necessarily the level of contribution of an incumbent technology that declines over time, it is the return on investment in technology improvements that declines. The question here is: which technologies should I invest in to reach the next level of product and service quality and reliability?

²Recently, Netflix has offered the Netflix Prize (http://www.netflixprize.com), a $1 million award to any person or organization that produces a movie recommendation algorithm 10% better than the existing Netflix algorithm (Ellenberg 2008). An Internet search on “netflix dataset” will provide the reader with interesting insights into this application of data mining and machine learning, a snapshot of how the Netflix Prize competitors are doing, and links to the actual Netflix data set.

³The reader is encouraged to look up the Claritas description of their own ZIP code at: http://www.clusterstaging.claritas.com/MyBestSegments/Default.jsp

⁴CRISP-DM is comparable to the DMAIC improvement cycle (Cios et al. 2007) commonly associated with Six Sigma approaches. DMAIC is composed of five phases: define, measure, analyze, improve, and control, which are generally comparable to the six CRISP-DM steps. Because KDAM approaches typically rely on real-time on-line production data gathered during regular process operation (vs. data obtained during designed and controlled experiments), the data can be quite noisy. Thus, the CRISP-DM model places significant emphasis on data understanding and preparation. Beyond this, examination of the details of the process steps reveals that differences among these and related methodologies often lie more in origins rather than intent, with approaches such as CRISP-DM being perhaps more closely associated with information technology and DMAIC and related methodologies having origins in engineering.

⁵Contact Dr. Andrzej Kochanski, Institute of Manufacturing Technologies, Warsaw University of Technology, Warsaw, Poland, [email protected]

⁶Contact Dr. Mark Polczynski.

⁷In addition to potentially facilitating analysis, normalization “anonymizes” data, thus easing concerns about security of sensitive process and quality/reliability information.

⁸If these cases at Toyota and Honda are indicative of an east-first spread of the application of data mining and machine learning technology to improving product quality and reliability, then the reader might take a moment to reflect on history's penchant for repeating itself.

⁹This univariate control chart is used for purposes of illustration. Latent-based multivariate control charts (Cios et al. 2007; Han and Kamber 2006; Kourti and MacGregor 1996; Lewicki and Hill 2007; MacGregor 1997; MacGregor and Kourti 1995) constitute a more sophisticated approach.

¹⁰When applying neural networks, a particular problem can typically be effectively solved by networks with different architectures, topologies, and network weights, raising the question: Which is the “right” network?