Fuzzy control charts for process quality improvement and product assessment in tip shear carpet industry

Purpose - The purpose of this paper is to demonstrate the use of artificial intelligence methods in quality control and improvement. The paper introduces a systematic approach for the design of fuzzy control charts of tip shear carpets. Design/methodology/approach - There are certain steps for designing fuzzy control charts. All input, state and output variables of the carpet plant and partition of the universe of discourse were first determined. The interval spanned by each variable and the number of fuzzy subsets each assigned with a linguistic label were identified. Then, the adaptive capability of neural network was used to determine the membership functions for each fuzzy subset. The fuzzy relationship functions between the inputs and outputs were assigned to form the fuzzy rule base (controller) in order to normalize the variables and certain intervals. Fuzzification of input parameters and max-min composition of rules for inferring crisp outputs was the next step. The aggregation of fuzzified outputs and defuzzification of the outputs were the last step of this study, which helped to produce crisp outputs for latex weight. Findings - Fuzzy linguistic terms were employed for overall quality assessment and rating of the end product. The outcomes of neuro-fuzzy system were good supplements to other statistical process control tools. Research limitations/implications - Lack of qualified domain experts, knowledge acquisition of process parameters and time limitation for training of neuro-fuzzy model were primary limitations. Practical implications - The approach is more flexible and meaningful to identify the quality distribution of a product. The qualitative aspect of human reasoning for decision making was employed in this approach. Originality/value - The paper is original and the first such work for local industry.