Apply artificial neural network to model high-pressure processing.

Most high-pressure food applications are not only pressure-dependent but also temperature-dependent, making temperature an important process parameter. Temperature distributed inside the pressure vessel needs to be as homogeneous as possible so that researchers can extract valid conclusions about the effects that a given treatment has on a specific food.

But making temperature uniform during the high-pressure processing of foods is not easy because the complex phenomena of heat and mass transfer are involved. Modeling high-pressure processes is a challenging task mainly because it is difficult to obtain information on the appropriate thermal properties of the materials under pressure. Because of this difficulty, different approaches have been attempted, including the use of a variety of mathematical models.

Researchers in Spain indicate that an artificial neural network could be used to obtain a temperature prediction model for high-pressure processes. The main advantage of such a network is its ability to learn from examples, without requiring a previous knowledge of the relationships that exist between the process parameters. It's an emulation of a biological neural system. A neural network can perform tasks that a linear program cannot. A neural network learns and does not need to be reprogrammed. The utility of artificial neural network models lies in the fact that they can be used to infer a function from observations and also to use it. This is particularly useful in applications where the complexity of the data or task makes the design of such a function by hand impractical.

In order to learn from previous examples, these neural networks require an important training phase. In most cases, the network can change its structure based on external or internal information that flows through the network during its training phase. To predict temperature, the models would require equations that describe all heat transfer phenomena that could occur during the compression, pressure-holding and decompression phases of the high-pressure process. While this type of model may be complicated to construct, it could give a complete and accurate vision of all the parameters involved and their relationships to each other.

Further information. Berengere Guignon, Department of Engineering, Instituto del Frio, C/Jose Antonio Novais, 10 Madrid E-28040 Spain; phone: [phone]; fax: [phone]; URL: www.if.csic.es.