2. Intelligence Systems/Soft Computing Systems in Clothing Comfort Evaluation

Neuro computing, fuzzy logic, and other evolutionary algorithms such as artificial neural network (ANN) and adaptive neuro-fuzzy inference systems (ANFIS) which are used for the prediction, modeling, and optimization of textile properties are called soft computing systems or intelligent systems ^[16][75]. Human tactile sensations for the textile goods are very complex ^[17][76] and hence linear modelling cannot solve such relationships. The comfort of textile goods, which is closely related to the user’s perception and acceptance, has been currently studied and relationships between the physiological perception and the instrumental data have been estimated using various intelligence techniques. The most frequently used intelligence methods applied in the prediction of the tactile comfort of textile goods are fuzzy logic, artificial neural network (ANN), and adaptive neuro-fuzzy inference system (ANFIS). Settle et al. ^[16][75] discussed the importance of soft computing/intelligence techniques in the textile industry. The use of fuzzy logic has been used by many researchers recently after LA Zadeh’s pioneering work in 1965 ^[18][77]. Zadeh applied fuzzy logic for solving uncertainty or fuzziness or ambiguity of different parameters. The comfort of textile goods, particularly physiological perception, is full of uncertainty and fuzzy logic was the perfect fit for this concept. After Zadeh’s work, Zimmerman ^[19][78] and Mamdani ^[20][79] introduced the IF-THEN rules which support the ideas of Zadeh. Thus, implementing fuzzy logic in the sensory evaluation of textile goods was started in 1991 by Raeel and co-workers ^[21][53]. The authors applied the fuzzy transformation matrix to predict the handle of lightweight dress fabrics and found good agreement between the subjective and the objective evaluation of the same fabric. The fuzzy transformation matrix enabled them to calculate the handle of the fabrics quantitatively, which further can be used as quality data which can be used for the communication between the user and the manufacturer. However, fuzzy logic has its own limitations as it does not have a specified technique that can be used as a mentor in the process of translating human perception into a rule-based fuzzy inference system (FIS) ^[22][80]. Other researchers ^{[23][24][25][26][27]}[81,82,83,84,85] have applied fuzzy logic to predict and represent the comfort of textile goods. The conclusions from all authors confirmed that fuzzy logic is a better method for comfort prediction of textile goods when compared to linear statistical methods. However, the hybrid models, adaptive neuro-fuzzy inference system (ANFIS), can perform better than the fuzzy logic model. This is because ANFIS uses the FIS system to transform human knowledge into the rule-based fuzzy inference system ^[28][86]. To observe the difference between the efficiency of fuzzy logic and ANFIS, we used the work performed by Jeguirim et al. ^[29][87] on the use of the fuzzy logic model to predict the handle of knitted fabric from process parameters. We took the data from their work and we performed ANFIS using Matlab2017b^® and found the following results (see Figure 210). A linear fit was employed to observe the difference between the two models (Figure 10). As observed, the linear model has a better fit for the ANFIS model. The values obtained by the fuzzy logic model were highly scattered. Hence, the ANFIS model performed better in predicting the handle of the textile goods. The root mean square error (RMSE) and mean relative percent error (MRPE) were calculated for ANFIS and compared with the fuzzy logic model from Jeguirim and his co-worker’s work. These two parameters were used to evaluate the prediction performance of the models. $$RMSE=\sqrt{\frac{1}{N}\sqrt{}{\displaystyle \sum }_{i=1\underset{}{\overset{}{n}}{(p-y)2^{}}^{}}^{}}$$ where N; total number of observations, y; actual values, p; predicted values. Based on this computation, the calculated RMSE, RMPE, and the standard deviation values for the ANFIS model were 0.083, 0.062, and 0.72, respectively, while the values according to the work of Jeguirim et al. were 0.29, 1.23, and 0.73, respectively. These results affirmed that the performance of the ANFIS model is higher than that of the fuzzy logic models. This is because fewer errors were observed in the case of the ANFIS model. Hence, this review paper recommends the ANFIS model over the fuzzy logic models. The adjusted R2 value is high. This confirms that the ANFIS model is the best model for predicting the handle of textile goods. A similar review was made by Zeng et al. ^[30][88] (Figure 311). The researchers addressed the use of fuzzy logic to integrate human perception with the instrumental data on textile goods. They applied principal component analysis to support the extraction of fuzzy rules so that they could build the hand evaluation model. They found that fuzzy logic is an excellent model for such hand integration and estimation. We performed the ANFIS algorithm using the inputs from the paper and compared the results with fuzzy logic. We obtained the following result.