为建立生长/非生长界面模型来预测溶藻弧菌在环境因子交互作用下的生长概率,探究环境因子对溶藻弧菌的生长的交互作用,该文选取轻腌大黄鱼室温25℃贮藏货架期终点分离的溶藻弧菌作为研究对象,研究25℃下pH值、水分活度(water activity,aw)和盐分对其生长概率的交互影响.使用Gompertz模型对其生长情况进行拟合,比较其生长动力学参数.用简单Logistic方程、二阶线性Logistic回归方程拟合及概率神经网络算法(probabilistic neural network,PNN)建立溶藻弧菌生长/非生长界面模型,使用正确率、假阳性率对其拟合优度进行比较.结果表明:二阶线性Logistic回归方程拟合效果更优,验证集的正确率为90.9%,PNN验证集正确率为90.0%.随着盐分的增大,生长/非生长的界限明显向低水分活度、低pH值方向移动;在相同盐分条件范围内,高水分活度且pH值较高条件下,比生长速率较高,延滞期也相应较短;随着盐分的增长,0.91与0.90低水分活度条件下溶藻弧菌也开始缓慢增长,但存在较长时间的延滞期,高盐分对溶藻弧菌有生长促进作用.研究表明:PNN可在工业生产中对溶藻弧菌的生长和非生长快速预测分类,通过二阶线性Logistic可评估pH值、aw和盐分许用范围内水产品的安全性.通过构建溶藻弧菌概率模型和动力学模型,可为改进贮藏条件和产品配方、确保轻腌大黄鱼质量安全提供支持.%The aim of the study was to develop a growth/no-growth interface model to predict the growth probability of Vibrio alginolyticus associated with lightly salted Pseudosciaena crocea under 3 environmental factors,and to explore the inhibitory effect of environmental factors on the growth kinetics of target micro-organism. The effects of pH value, water activity (aw) and NaCl content on the growth probability of the Vibrio alginolytica were studied at ambient temperature(25 ℃).At present, mainly the "fence technology" is used to change the growth environment of microorganisms by changing the water activity, salt, acetic acid, Nisin and sugar, so as to achieve the role of inhibition to microorganisms. Logistic regression is a commonly used method for simulating the microbial growth boundary (growth/non-growth interface) in food, through which the growth environment can be adjusted and the shelf life can be extended. Artificial neural network model PNN (probabilistic neural network) is a feed forward neural network with strong nonlinear pattern classification ability and high accuracy of nonlinear algorithm, which can solve the growth/non-growth interface problems, and the PNN has simple structure and high training speed without considering the complex chemical reaction during storage. Simple logistic equation, second-order linear logistic regression equation and PNN artificial neural network model were used to establish the growth/non-growth interface model of Vibrio alginolyticus,while fraction correct(FC)and false alarm rate(FAR)were used to compare the goodness of fit of the 3 models. The Gompertz model was used to fit the growth condition, and the growth kinetics parameters were obtained. The results showed that the second-order linear logistic regression equation had better fitting results, the consistency index of the training set was 94.8%, and that of the validation set was 90.9%, while the consistency index of the PNN artificial neural network was 95.6% and 90.0% for the training and validation set, respectively. The FAR of the second-order linear logistic regression equation was 5% (training set) and 0 (validation set), while that of the PNN artificial neural network was 6.6% (training set) and 22% (validation set). The effects of the environmental factors were as follows: With the increase of salt content, the growth/no-growth boundary obviously moved to low water activity and low pH value. In the same salty condition, in the range of high aw and high pH value, the growth rate was higher and the retardation period was shorter. With the increase of salt content,even under low aw such as 0.91 and 0.90,the Vibrio alginolyticus also began to grow slowly,but there was a long lag time. The conclusions are obtained: PNN artificial neural network can do quick classification prediction on the growth/no-growth data of Vibrio alginolyticus in the industrial production,and the second-order linear logistic regression can evaluate the stability of aquatic products under the conditions of aw, pH value and salt content. By constructing the probabilistic models and kinetic models of Vibrio alginolyticus which can assess the stability of characteristic aquatic products in the range of pH value, aw and salt content, it can provide the guide to suppress microorganisms without the use of chemical preservatives to ensure quality and safety of pickled Pseudosciaena crocea.
展开▼
