Asnaashari Maryam, Farhoosh Reza, Farahmandfar Reza
Department of Food Science and Technology, Sari Agricultural Sciences & Natural Resources University (SANRU), P.O. Box 578, Sari, Iran.
Department of Food Science and Technology, Faculty of Agriculture, Ferdowsi University of Mashhad, P.O. Box 91775-1163, Mashhad, Iran.
J Sci Food Agric. 2016 Oct;96(13):4594-602. doi: 10.1002/jsfa.7677. Epub 2016 Mar 22.
As a result of concerns regarding possible health hazards of synthetic antioxidants, gallic acid and methyl gallate may be introduced as natural antioxidants to improve oxidative stability of marine oil. Since conventional modelling could not predict the oxidative parameters precisely, artificial neural network (ANN) and neuro-fuzzy inference system (ANFIS) modelling with three inputs, including type of antioxidant (gallic acid and methyl gallate), temperature (35, 45 and 55 °C) and concentration (0, 200, 400, 800 and 1600 mg L(-1) ) and four outputs containing induction period (IP), slope of initial stage of oxidation curve (k1 ) and slope of propagation stage of oxidation curve (k2 ) and peroxide value at the IP (PVIP ) were performed to predict the oxidation parameters of Kilka oil triacylglycerols and were compared to multiple linear regression (MLR).
The results showed ANFIS was the best model with high coefficient of determination (R(2) = 0.99, 0.99, 0.92 and 0.77 for IP, k1 , k2 and PVIP , respectively). So, the RMSE and MAE values for IP were 7.49 and 4.92 in ANFIS model. However, they were to be 15.95 and 10.88 and 34.14 and 3.60 for the best MLP structure and MLR, respectively. So, MLR showed the minimum accuracy among the constructed models.
Sensitivity analysis based on the ANFIS model suggested a high sensitivity of oxidation parameters, particularly the induction period on concentrations of gallic acid and methyl gallate due to their high antioxidant activity to retard oil oxidation and enhanced Kilka oil shelf life. © 2016 Society of Chemical Industry.
由于担心合成抗氧化剂可能存在的健康危害，没食子酸和没食子酸甲酯可作为天然抗氧化剂引入，以提高海产油的氧化稳定性。由于传统建模无法精确预测氧化参数，因此进行了人工神经网络（ANN）和神经模糊推理系统（ANFIS）建模，其具有三个输入，包括抗氧化剂类型（没食子酸和没食子酸甲酯）、温度（35、45和55°C）和浓度（0、200、400、800和1600 mg L⁻¹），以及四个输出，包含诱导期（IP）、氧化曲线初始阶段的斜率（k1）和氧化曲线传播阶段的斜率（k2）以及IP时的过氧化值（PVIP），以预测基尔卡油三酰甘油的氧化参数，并与多元线性回归（MLR）进行比较。
结果表明，ANFIS是最佳模型，其决定系数较高（IP、k1、k2和PVIP的R²分别为0.99、0.99、0.92和0.77）。因此，ANFIS模型中IP的RMSE和MAE值分别为7.49和4.92。然而，最佳MLP结构和MLR的RMSE和MAE值分别为15.95和10.88以及34.14和3.60。因此，MLR在所构建的模型中显示出最低的准确性。
基于ANFIS模型的敏感性分析表明，氧化参数具有较高的敏感性，特别是诱导期对没食子酸和没食子酸甲酯浓度的敏感性，这是由于它们具有较高的抗氧化活性来延缓油氧化并延长基尔卡油的保质期。©2016化学工业协会。
