High pressure processing (HPP) is a non-thermal food preservation method that inactivates microorganisms and enzymes by applying high pressure, typically up to 600 MPa (Georget et al., 2015; Terefe, Tepper, Ullman, Knoerzer, & Juliano, 2016). The process is known to cause an adiabatic temperature increase in the treated product, a phenomenon known as compression heating (Knoerzer, Juliano, Gladman, Versteeg, & Fryer, 2007). This is a critical factor in the design and optimisation of HPP treatments, and particularly in high-pressure thermal processing (HPTP), as it can significantly influence the inactivation kinetics of microorganisms (Gänzle, 2023).

HPTP, which operates at pressures of up to 600 MPa and temperatures (under pressure) ranging from 60 to above 120 °C, has been hailed as a disruptive technology in food processing for well over a decade. Significant research by various groups has investigated aspects such as bacterial spore inactivation (since spores are generally not affected by high pressure alone) (Setlow & Doona, 2023) and food texture modification (Sikes, 2023; Terefe, Kumkanokrat, Limos, & Bergkvist, 2023). The potential of the technology was recognised; however, the lack of available commercial-scale equipment reduced momentum for continued investigation and optimisation of processes and equipment. Recent developments in utilising a drop-in solution to allow for HPTP in conventional machines (commercially available as of mid-2023; (Knoerzer, 2017)) have led to a resurgence in HPTP interest in both academia and industry. Quantitatively understanding, accurately predicting, and optimising the extent of compression heating for any material involved in a HPTP process is essential, especially when considering insulating materials that could be used in HPTP.

Previously, we developed a method for determining the pressure- and temperature-dependent compression heating properties of liquids, specifically water and propylene glycol, and their mixtures (Knoerzer, Buckow, Sanguansri, & Versteeg, 2010). This method involved recording temperature profiles in the samples during high-pressure processing, commencing at a wide range of initial temperatures. The temperature profiles were then analysed with an algorithm to determine the compression heating properties as functions of pressure and temperature. This data was then used to solve an ordinary differential equation (ODE, Eq. (1), see below) which predicts the compression heating of these materials under high pressure conditions.

In a subsequent study, we extended this approach to solid materials, specifically polytetrafluoroethylene (PTFE), polypropylene (PP), and high-density polyethylene (HDPE) (Knoerzer, Buckow, & Versteeg, 2010). These materials are often used in HPP as components of the pressure vessel or product packaging, and their compression heating behaviour can significantly affect the temperature distribution within the vessel during processing. Therefore, understanding and then being able to model, the behaviour of these materials during HPP, and more importantly HPTP, is crucial for ensuring food safety and quality.

In this paper, we focus on the compression heating behaviour of water, propylene glycol (PG), PP, HDPE, and PTFE. We will utilise the same experimental data from our previous studies but apply a new approach – artificial neural networks (ANNs). An ANN is a computational model inspired by the structure and function of biological neural networks, consisting of interconnected nodes or neurons, organised into layers. ANNs are capable of approximating complex nonlinear relationships and are widely used in various fields (Ibrahim, Elsheikh, Elasyed Abd Elaziz, & Al-Qaness, 2022).

In recent years, ANNs have gained prominence in the food science sector, finding applications in quality control, safety monitoring, product development, and consumer behaviour analysis (Bhagya Raj & Dash, 2020; Ma et al., 2022; Nayak, Vakula, Dinesh, Naik, & Pelusi, 2020). Their ability to model complex, non-linear relationships has proven invaluable in tackling the intricate challenges presented in food process engineering, including food safety and quality analyses, food image analysis, and modelling of various thermal and non-thermal processes. Furthermore, the use of ANNs has been identified as a pivotal opportunity in safeguarding the integrity and safety of the global food supply chain, with a significant impact in agricultural markets. This wide-ranging applicability and potential for innovation underscore the relevance of employing ANNs in our study, providing a promising alternative to traditional modelling methods.

Here, we employ ANNs to predict adiabatic compression heating, leveraging their ability to capture intricate patterns in the data. It is our expectation that this new approach is less labour-intensive and time-consuming compared to the previous (ODE) method and has potential to provide more accurate and reliable predictions. By comparing the results of this ANN approach with those of the previous approach, we will be able to assess the relative strengths and weaknesses of both methods and provide guidance for future research in this area.