---

Construction of buildings and structures causes to compact of soil particles and soil settlement. Hence, determination and prediction of soil settlement in the stability of structures, resulting from the applied loads, is necessary before construction. As a result of consolidation test that is relatively time-consuming and costly testing, compression index (Cc) is used to get the amount of settlement. In fact, soil settlement can cause extensive damage to a project in some cases. In order To prevent these damages, correct prediction can be useful for safe designing of structures. Cc may be as a function of various parameters such as initial void ratio of soil, moisture of liquid limit, moisture of plastic limit, plasticity index, relative density, and so on. By considering the longtime of consolidation test, researchers have tried to find relationship between these parameters and Cc from the past until now. For this reason they tried to connect Cc to other physical measurable properties of the soil.

In the past, some researchers have indirectly tried to measure these parameters. In this regard, several empirical single-parameter approaches are proposed to estimate Cc. Due to non-linear relationship between Cc and relevant parameters, Adaptive Neuro-Fuzzy Inference System (ANFIS) has found as an application to solve such non-linear problems and cases where an accurate understanding of the problem is required. ANFIS is a multilayer feed forward networks that is combination of Fuzzy Inference System (FIS) and Neural Network (NN). NN has ability to learn the input and output data and FIS is also capable for map the input space to the output space. ANFIS is a powerful tool to solve complex and nonlinear problems using the two mentioned features and also language power of FIS and numerical power of adaptive nervous system.

In this paper, Compression index (Cc) is modeled by ANFIS. Two ANFIS model were created by subtractive clustering (SC) and Fuzzy c-means clustering (FCM), respectively, and then trained. By data clustering, collection of training data is divided into a number of fuzzy clusters and each cluster representing the system behavior. The data were collected from the Soil Mechanics Laboratory of Mashhad city. ANFIS input parameters are taken according to the same parameters that commonly chosen in most of empirical models for determining Cc that easily determined in the laboratory. These input parameters include initial void ratio (e0), liquid limit (LL) and plastic limit (PL).
The number of required iterations for training (Epochs) in two ANFIS model, neighborhood radius (ra) in SC and number of clusters (NC) in FCM are optimized using trial and error method. After the end of solving and optimization of ANFIS models, the SC-FIS model was found in ra = 0.25 and NC =18 and the FCM-FIS model was obtained in NC = 20 with highest accuracy for prediction. Results showed both ANFIS model have a high capacity and appropriate forecasting for Cc prediction with chosen inputs parameters. Compared to the SC-FIS model, FCM-FIS is conducted prediction with higher accuracy. Using presented ANFIS models, can be predict the Cc of soils whose characteristics are within the specifications soils that used in this modeling with high accuracy and do not need to conduct consolidation tests that are very time consuming and costly.

---

Adaptive neuro-fuzzy inference system (neuro-fuzzy or ANFIS) is a well-known hybrid neuro-fuzzy network for modeling complex systems. In this system ,the most frequently used fuzzy clustering method is the fuzzy subtractive clustering algorithm. In this algorithm, a cluster with a certain degree has each data point, explained by a membership function level. In this study, ANFIS model was used for prediction of weight reduction (%), solid gain (%),water loss (%) and rehydration (%) of banana slices dehydrated by osmosis-ultrasound method. The ANFIS model was developed with 3 inputs of sonication power (at three levels of 0, 75 and 150 watts), ultrasound treatment time (at three times of 10, 15 and 20 minutes) and sucrose solution concentration (at three levels of 30, 45 and 60 °Brix) to predict the characteristics of dehydrated banana slices. The calculated coefficient of determination values for prediction of weight reduction (%), solid gain (%),water loss (%) and rehydration (%) of dehydrated banana slices using the ANFIS-based subtractive clustering algorithm were 0.93, 0.95, 0.94, and 0.91, respectively. In general, it can be said that the high coefficients of determination between the experimental results and the outputs of the ANFIS model indicate acceptable accuracy and usability this method in controlling complex processes in the food industry, including dehydration and drying processes.

---

ANFIS (Adaptive neuro-fuzzy inference system) is a combined neuro-fuzzy method for modeling transport phenomena (mass and heat) in the food processing. In this study, first, an infrared dryer was used to dry the extracted gum from quince seed. Then, ANFIS method was used to modeling and predicting the weight changes percentage of this gum when drying in infrared dryer. In the infrared dryer, the effect of samples distance from the radiation lamp and the effect of the gum thickness inside the container on the drying time and the weight loss percentage of quince seed gum during drying time were investigated. The results of drying of this gum by infrared method showed that by reducing the samples distance from the heat source from 10 to 5 cm, the average drying time of quince seed gum decreased from 58.0 minutes to 29.3 minutes (thickness 1.5 cm). Also, by reducing the gum thickness in the sample container from 1.5 to 0.5 cm, the average drying time of the extracted gum decreased from 45.7 minutes to 19.3 minutes (distance 7.5 cm). The ANFIS model was developed with 3 inputs of drying time, samples distance from heat source and gum thickness in the sample container to predict the weight changes percentage of this gum when drying in infrared dryer. The calculated coefficients of determination values for predicting the weight loss percentage of gum using the ANFIS-based subtractive clustering algorithm was 0.983. In general, it can be said that the high coefficients of determination between the experimental results and the outputs of the ANFIS model indicate the acceptable accuracy and usability of this method in modeling heat and mass transfer processes in the food industry.