This study uses artificial neural network analysis (ANN) to predict quality properties of dried cranberries. Drying techniques were used including osmotic pretreatments (with or without ultrasound) in combination with hot-air and microwave assisted hot-air. To carry out osmotic process, a ternary solution (Sucrose-Nacl- Water) was used. This model mathematically correlates five processing variables (concentration of sucrose, concentration of Nacl, temperature of osmotic solution, frequency of ultrasound (0, 35,130 kHz) and power of microwave (0, 180 and 300 W)) with color, water activity, texture, total anthocyanins, polymeric color and antioxidant activity. Quality properties of dried cranberries were examined by a Hunterlab Colorimeter for color, a_w meter for water activity, Texture analyser for texture, spectrophotometer for total anthocyanins, polymeric color. Antioxidant activity was measured using EC_(50) method, too. Texture of f ultrasonic (130 kHz) pretreated samples when dried with MW (300 W), was similar to samples dried with hot-air. When using microwave, steam pressure caused severe collapse of textural layers in frequency of 130 kHz that it was because of severe destruction created in this frequency. Therefore, higher maximum force (N) was seen. Optimized ANN models were developed based on 10-20 neurons per hidden layer. ANN models were then tested against an independent dataset. The optimal ANN consisted of 2 hidden layers with 18 neurons. Measured values of outputs were predicted with an R~2>0.87.
展开▼
