A machine learning-based framework for automatic identification of process and product fingerprints for smart manufacturing systems. (January 2022)

• Record Type: Journal Article
• Authors: Kundu, Pradeep; Luo, Xichun; Qin, Yi; Cai, Yukui; Liu, Zhanqiang
• Is Part Of: Journal of manufacturing processes. Volume 73(2022)
• Page Start: 128
• Page End: 138
• Publication Date: 2022-01
• Subjects: Smart manufacturing -- Quality control -- Machine learning -- Process fingerprint -- Product fingerprint; Production management -- Data processing -- Periodicals; Manufacturing processes -- Periodicals; Procestechnologie; Productietechniek; Production -- Gestion -- Informatique -- Périodiques; Fabrication -- Périodiques; Manufacturing processes; Production management -- Data processing; Periodicals; 670.5
• DOI: 10.1016/j.jmapro.2021.10.060
• Languages: English
• ISSNs: 1526-6125

Abstract

Abstract: Process and product fingerprints are unique process and product characterisation parameters, respectively. They need to be controlled to ensure that manufactured parts are within their specifications to achieve the required functions. The process and product fingerprints are developed to reduce production time and efforts for metrology and process optimisation. The existing methods to identify process and product fingerprints are usually developed for particular machining processes. They are often performed manually which make them time-consuming. In addition, the existing methods are strongly influenced by human decisions. As a result, the fingerprints identified by these methods may have poor correlations with product functionality or the surface quality/dimension characteristics. For the first time, a unique machine learning-based framework is presented in this paper to automatically identify process and product fingerprints. The new framework can shorten fingerprint development time and reduce complexity of process control and product characterisation. A CTER index which is the ratio of correlation and testing error ratio has been proposed to determine the best fingerprint in the framework. The performance of the proposed framework has been validated using data obtained from manufacturing superhydrophobic structures on stainless steel using a nanosecond laser. The correlation of product functional characteristics with fingerprints identified using the proposed framework is observed better than existing fingerprints developed purely based on the physics of the machining process. In addition, the proposed framework can also be used for process optimisation and prediction of functional characteristics/measurement tolerances, which has also been presented in this paper. Highlights: A unique machine learning-based framework has been presented to automatically identify process and product fingerprints. A correlation to testing error ratio (CTER) index has been proposed for finding the best fingerprint in the framework. Proposed framework performance has been validated using data obtained from manufacturing superhydrophobic structures. High correlation between functional characteristics and fingerprints identified using the proposed framework is observed.