Prediction of food thermal process evaluation parameters using neural networks

Two neural networks (ANN) were developed to predict thermal process evaluation parameters g and f(h)/U (the ratio of heating rate index to the sterilizing value), respectively. The temperature change required for the thermal destruction curve to traverse one log cycle (z), cooling lag factor (j(c)) andf(h)/U were input variables for predicting g and z, while j(c) and g were inputs for predicting f(h)/U. The data used to train and verify the ANN were obtained from reported values. Shrinking of input and output variables using natural logarithm function improved the prediction accuracy. The use of "Wardnets" with three slabs of 14 nodes in each slab, with a learning rate of 0.7 and momentum of 0.9 provided the best predictions. The g (unshrunk values) was predicted with a mean relative error of 1.25 +/- 1.77%, and a mean absolute error of 0.11 +/- 0.16 degrees F. The f(h)/U was predicted with a mean relative error of 1.41 +/- 3.40%, and a mean absolute error of 2.43 +/- 15.97, using 10 nodes in each slab. The process time calculated using the g from the ANN models closely followed the time calculated from the tabulated gvalues (RMS=0.612 min, average absolute error=0.466 min with an S.D. of 0.400 min).     

Prediction of moisture in cheese of commercial production using neural networks

Control of cheese moisture is paramount to maximizing yield and profitability of a cheesemaking operation. Modeling and prediction of cheese moisture prior to pressing from a large industrial database for stirred-curd Cheddar cheese made with non-standardized and standardized milk was carried out using neural networks (NN). The number of model input variables was reduced by removing or combining some of them, based on cheesemaking knowledge and on the results of two tests estimating the impact of each model input. Input removal was carried out until the validation mean absolute prediction error (MAPE) increased. An initial NN cheese moisture model with 38 input process variables, coded as 57 NN inputs, was reduced to one with 21 input process variables, coded as 34 NN inputs. For the latter, the validation MAPE was 0.53% cheese moisture in a range of cheese moisture of 13.2%, and 0.51% for the best 25% of models (out of 100). For the range of operating conditions of the process in this study, four main groups of variables were found to be the most influential on the prediction of cheese moisture: cutting and subsequent stirring of the curd, curd rinsing temperature, starter quantity, activity and strain, and seasonal variation of milk composition. The NN model with the selected input variables and optimized number of hidden neurons was then used to predict cheese moisture for ranges of these variables. This study showed that NN models can successfully extract input–output variable relationships from industrial production data in spite of the inherent error in these data. The resulting NN models can be used both for research to develop the base of knowledge on production variables and their complex interactions, as well as for the prediction of cheese moisture.     

Prediction of wheat quality parameters using near-infrared spectroscopy and artificial neural networks

In wheat and flour processing, the quality control needs quick analytical tools for predicting physical, rheological, and chemical properties. In this study, near infrared reflectance (NIR) spectroscopy combined with artificial neural network (ANN) was used to predict the flour quality parameters that are protein content, moisture content, Zeleny sedimentation, water absorption, dough development time, dough stability time, degree of dough softening, tenacity (P), extensibility (L), P/G, strength, and baking test (loaf volume and loaf weight). A total of 79 flour samples of different wheat varieties grown in different regions of Turkey were chemically analyzed, and the results of both NIR spectrum (400–2,498 nm) and chemical analysis were used to train/test the network by applying various ANN architectures. Prediction of protein, P, P/G, moisture content, Zeleny sedimentation, and water absorption in particular gave a very good accuracy with coefficient of determination (R ²) of 0.952, 0.948, 0.933, 0.920, 0.917, and 0.832, respectively. The results indicate that NIR combined with the ANN can successfully be used to predict the quality parameters of wheat flour.     

Prediction of mass transfer kinetics during osmotic dehydration of apples using neural networks

An artificial neural network (ANN) model was developed for the prediction of water loss (WL) and solid gain (SG) in osmotic dehydration of apple cylinders using a wide variety of data from the literature to make it more general. This model mathematically correlates six processing variables (temperature and concentration of osmotic solution, immersion time, surface area, solution to fruit mass ratio and agitation level) with WL and SG. The optimal ANN consisted of one hidden layer with four neurons. This model was able to predict WL and SG in a wide range of processing variables with a mean square error of 13.9 and 4.4, and regression coefficient of 0.96 and 0.89, respectively, in testing step. This ANN model performs better when compared to linear multi-variable regression. The wide range of processing variables considered for the formulation of this model, and its easy implementation in a spreadsheet using a set of equations, make it very useful and practical for process design and control.     

Prediction of texture characteristics from extrusion food surface images using a computer vision system and artificial neural networks

Surface images and the texture characteristics of 17 samples and the 25 different parts within one sample were detected using a computer vision system and texture profile analysis in extruded food. According to the linear fitting model, the hardness and gumminess score can be reflected directly by the a^* and Intensity based on correlation coefficient of 0.9558, 0.9741 and 0.9429, 0.9619, respectively. The springiness could be reflected from color values through calculating from hardness and gumminess scores, indirectly. Neither of cohesiveness and chewiness presented relationship with two different color spaces. A desirable and accurate two hidden layers of back-propagation artificial neural network was trained for simulating and predicting the hardness and gumminess scores from a^* and Intensity based on the data in 17 samples, respectively. The simulation processing in ANN showed higher correlation coefficient of 0.9671 and 0.9856 than linear fitting model.     

小桐子油及其生物柴油多参数的物理特性预测及分析

运用相关物性数据估算方法,得到了小桐子油及其生物柴油的基本物性参数,有效地解决了临界参数难以测定的问题;同时估算出包括气体黏度、液体黏度、表面张力等与传递性质和平衡性质有关的物性参数;分别测定了不同温度下小桐子油及其生物柴油的相关物性,并与估算结果进行对比。结果表明:在实验温度下,小桐子生物柴油运动黏度的最大误差为5. 94%,表面张力的最大误差为5. 70%,估算结果较为准确。     

Prediction of consumer liking from trained sensory panel information: Evaluation of neural networks

This study was set out to establish artificial neural networks (ANN) as an alternative to regression methods (multiple linear, principal component and partial least squares regression) to predict consumer liking from trained sensory panel data. The sensory profile and acceptability of 10 market samples of beef bouillon products were measured. The products were distinct as evaluated by the trained sensory panel. A total of 100 regular beef bouillon product users from Manila measured overall liking, flavour, aftertaste and mouthfeel of the products. Curve fitting method was applied to identify sensory drivers of consumer liking. The sensory drivers of consumer liking were used as explanatory variables in artificial neural networks and regression methods. To overcome the limitations of regression methods we have used artificial neural network techniques to model consumer liking score as a function of trained sensory panel scores and achieved quite encouraging results. Our simulation experiments show that though the regression methods such as multiple linear regression (MLR), principal component regression (PCR) and partial least square (PLS) give an accurate prediction of consumer liking scores, this approach is not robust enough to handle the variations normally encountered in trained sensory panel data. ANNs were trained using the sensory panel raw data and transformed data. The networks trained with sensory panel raw data achieved 98% correct learning, the testing was in a range of 28–35%. Suitable transformation method was applied to reduce the variations in trained sensory panel raw data. The networks trained with transformed sensory panel data achieved about 80–90% correct learning and 80–95% correct testing. It is shown that due to its excellent noise tolerance property and ability to predict more than one type of consumer liking using a single model, the ANN approach promises to be an effective modelling tool.     

神经网络在混沌系统控制中的应用研究进展

神经网络在混沌系统控制中主要用于建模辨识或用作控制器，这2种作用都是以神经网络的任意非线性连续函数逼近能力为基础的。用神经网络进行建模辨识有2种途径：使用神经网络直接构造混沌系统的输入输出模型或通过神经网络构造混沌系统中的特征参量。神经网络应用于混沌系统建模辨识的研究主要集中在如何改进神经网络的学习算法和结构方面；神经网络用作控制器时主要用来解决混沌系统的轨迹跟踪或同步控制问题，其研究主要集中在改进控制结构与方法上，将神经网络应用于混沌系统控制已经取得了一些进展：动态神经网络开始应用于混沌系统辨识和控制；基于神经网络工作原理的函数网络能更容易地实现混沌系统的逼近。神经网络应用于混沌控制也存在一些需要解决的问题，如神经网络选择问题、神经网络结构问题、计算方法问题、硬件实现问题等。     

Spectroscopic quantification of bacteria using artificial neural networks

Fourier transform-infrared spectroscopy, in conjunction with artificial neural networks, has been used for identification and classification of selected foodborne pathogens. Five bacterial species (Enterococcus faecium, Salmonella Enteritidis, Bacillus cereus, Yersinia enterocolitica, Shigella boydii) and five Escherichia coli strains (O103, O55, O121, O30, O26) suspended in phosphate-buffered saline were enumerated to provide seven different concentrations ranging from 10(9) to 10(3) CFU/ ml. The trained artificial neural networks were then validated with an independent subset of samples and compared with the traditional plate count method. It was found that the concentration-based classification of the species was 100% correct and the strain-based classification was 90 to 100% accurate.     

Characterization of radial bending properties of cork

The behaviour of cork under three point bending stress in the radial direction was evaluated in relation to porosity (range 2.0?C15.4%) and density (range 0.160?C0.220?g?cm^?3). The study was made using water-boiled cork planks of two commercial quality classes (good and poor quality) collected at an industrial mill. Cork samples were cut with the largest dimension in two directions, axial and tangential, and with the load zone in the inner part and the outer part of the plank. The stress-strain curves obtained in bending were similar for the different test specimens and similar to those observed for tensile tests in cork but the mechanical resistance of cork in bending was higher than in tension and lower than in compression. The direction of the internal stress in samples submitted to bending was a highly significant factor of variation. The curves showed an initial linear elastic region with a mean Young??s modulus of 14?MPa for the tangential direction and 21?MPa for the axial direction, followed by a region of a gradual cork yielding up to a peak load, and failure at an average stress of 1.2?MPa and a strain of 14%. There was significant difference in the bending properties of cork samples obtained from cork planks of different quality classes. Density and porosity were not individually well correlated with the mechanical parameters. It was however possible to model Young??s modulus by combining porosity and density.     

BP神经网络在纺织工业中的应用

介绍了神经网络的发展历史,对BP网络的算法进行了详细的讨论,分析了BP网络在纺织工业中应用的主要步骤。     

An investigation on yarn engineering using artificial neural networks

Abstract

Engineering of spun yarns having specific tensile, evenness and hairiness characteristics is a long-cherished dream of spinning technologists. Selection of suitable raw materials at minimum cost and optimisation of process parameters are the two major tasks to be achieved to manufacture engineered yarn. Advent of high-speed fibre-testing machines and development of powerful modelling tools such as artificial neural network (ANN) have provided a great impetus in the yarn engineering research. This article demonstrates the feasibility of yarn engineering by developing a yarn-to-fibre ‘reverse’ model, using ANN. This approach is entirely different from the prevailing forward models, which predict the properties of final yarn using the fibre properties as inputs. The cost minimisation of cotton fibre mix was ensured by using the classical linear programming approach in combination with ANN. The engineered yarns demonstrated good agreement with the target yarn properties.     

Sorption and diffusion properties of volatile phenols into cork

The sorption and diffusion properties of seven common volatile phenols in hydro-alcoholic medium placed in contact with natural cork were investigated to determine the influence of cork closures on the concentration of these compounds in wine. Weighted cork samples were immersed in model wine solutions containing selected concentrations of each compound and were sampled over time. Sorption coefficients ranged between 125 and 306 mg of aroma compound per kg of cork, while diffusion coefficients varied from 0.8 to 4.1 × 10⁻¹¹ m²/s. Sorption isotherms of guaiacol and 4-propylguaiacol, respectively, the lowest and highest sorbed compounds of an homologous series of guaiacol, accurately fit Henry’s model. Hydrophobicity of aroma compounds seems to be a determining factor in both diffusion and sorption. This could indicate a preferential pathway for diffusion through suberin, the more hydrophobic constituent of cork.     

基于BP神经网络的精纺毛织物透气性能的预测

在34种精纺毛型织物实验数据基础上,利用三层BP神经网络方法,建立了织物透气性能与织物结构参数之间的神经网络模型,重新采集7种织物对网络模型进行验证和评估,结果表明:神经网络可以用来预测织物的透气性能指标,织物透气量的预测误差率范围为3%～24.2%,平均误差率为14.3%,最大误差率小于25%,神经网络预测精度受样本大小影响,进一步丰富学习样本后,神经网络的泛化能力可望得到改善.     

Prediction of freezing time for food products using a neural network

An artificial neural network (ANN) was developed to predict the freezing time of food products of any shape. The Pham model was used to generate freezing time data and to train ANN based on Wardnets. The product thickness (a), width (b), length (c), convective heat transfer coefficient (h_c), thermal conductivity of frozen product (k), product density (ρ), specific heat of unfrozen product (C_pu), moisture content of the product (m), initial product temperature (Ti), and ambient temperature (T_∞) were taken as input variables of the ANN to predict freezing time. The effects of the number of hidden layer nodes, learning rate, momentum on prediction accuracy were analyzed. The performance of the ANN was checked using experimental data. Predicted freezing time using the ANN was proved a simple, convenient and accurate method. Selection of hidden nodes, learning rate and momentum were important to ANN predictions.     

基于计算机视觉和人工神经网络预测挤压食品的质构特征

应用计算机视觉系统分别提取不同配方的挤压食品和同一样品不同部位的颜色值(HSI和L*、a*、b*),同时用质构分析仪测定样品质构特征。借助线性拟合模型通过样品的颜色对挤压食品的质构特征进行相关性分析,并利用BP神经网络模型通过颜色预测挤压食品的质构。线性拟合模型显示,硬度和胶粘度分别与a*值和对比度之间高度相关。两组实验中硬度与a*值之间的R2分别为0.9558、0.9429;胶粘度与对比度之间的R2分别为0.9741、0.9619。弹性与a*值和对比度之间具有一定的相关性,两组实验中弹性与a*值和对比度之间的R2分别为0.8675和0.8320。利用实验所得硬度、胶粘度、a*值以及对比度数据优化含有2个隐层的BP神经网络,得到两组实验对应最优网络模型结构,即每层所含神经元的数量分别为20、20,均方根(RMS,%)为4.25;20、40,均方根(RMS,%)为3.85。利用最优神经网络运用a*值和对比度对两组实验中的硬度和胶粘度进行模拟,得到的相关系数高于线性拟合模型拟合结果,两组实验中硬度与a*值之间的R2分别为0.9671、0.9770;胶粘度与对比度之间的R2分别为0.9766和0.9856。采用最优网络模型用颜色信息对挤压食品硬度和胶粘度的预测和验证结果表明,利用计算机视觉系统所提取的颜色值可以通过人工神经网络快速准确预测挤压食品的质构特征。      

Prediction of chemical,physical and sensory data from process parameters for frozen cod using multivariate analysis

Physical, chemical and sensory quality parameters were determined for 115 cod (Gadus morhua) samples stored under varying frozen storage conditions. Five different process parameters (period of frozen storage, frozen storage temperature, place of catch, season for catching and state of rigor) were varied systematically at two levels. The data obtained were evaluated using the multivariate methods, principal component analysis (PCA) and partial least squares (PLS) regression. The PCA models were used to identify which process parameters were actually most important for the quality of the frozen cod. PLS models that were able to predict the physical, chemical and sensory quality parameters from the process parameters of the frozen raw material were generated. The prediction abilities of the PLS models were good enough to give reasonable results even when the process parameters were characterised by ones and zeroes only. These results illustrate the application of multivariate analysis as an effective strategy for improving the quality of frozen fish products. © 1998 Society of Chemical Industry.     

Modifying the physical properties of butter using high-intensity ultrasound

The physical qualities of butter are affected by the physical properties of the cream used to make it. The objective of this study was to evaluate the effect of high-intensity ultrasound (HIU) on the physical properties of cream and butter. High-intensity ultrasound (frequency: 20 kHz, amplitude: 108 µm), often called sonication, was applied for 0, 10, 30, 60, and 90 s using a 1.27-cm-diameter tip to heavy cream (40% fat; 300 g) that was aged at 7.5°C with low agitation (40 rpm) for 90 min. Sonicated cream was churned at 7.5°C until butter grains were formed. The solid fat content (SFC), melting behavior, and average fat droplet size of cream were measured after HIU treatment. Butter was characterized by SFC and melting behavior immediately after production and was tested for SFC, melting behavior, and hardness after storage for 24 h at 5°C. High-intensity ultrasound did not affect the average fat droplet size of cream. Sonicating cream for 30, 60, and 90 s slightly decreased SFC due to the temperature increase (2–6°C) that occurred during HIU application. Two melting peaks were observed at approximately 17 and 33°C in all the cream samples. A significantly lower peak temperature was observed in cream sonicated for 10 s compared with creams sonicated for 30 and 60 s. A relatively shorter churning time of sonicated cream compared with nonsonicated cream was observed, possibly because HIU weakens the fat globule membrane. Two melting peaks were observed in all butter samples at approximately 16 and 33°C. Treatment with HIU for 10 to 60 s significantly increased the hardness of butter. When HIU was applied to cream for 10 s, the hardest butter was obtained, with the lowest onset temperatures and highest enthalpy values for both melting peaks. Treatment with HIU for 10 s promoted crystallization of low-melting-point triacylglycerols (TAG) during churning, which resulted in a harder material. Significantly lower enthalpy values for the high-melting-fraction TAG were observed in butters treated with HIU for 60 and 90 s compared with non-HIU-treated butter, which suggests that a longer duration of HIU promotes melting of high-melting-point TAG. The hardness of butter was correlated with the enthalpy values of the low-melting fraction and with total enthalpy values of fresh butter. However, further crystallization occurred in the butter during 24 h of storage at 5°C, and all differences in enthalpy values disappeared. In conclusion, exposure of cream to HIU can be used to modify the physical properties of butter, and the effects of HIU depend on the length of HIU treatment.