Process regime classification and modelling of a sequencing batch reactor for producing polyhydroxybutyrate with mixed culture using neural networks

Abstract

In recent years, there has been a growing incentive towards production and application of environmentally benign materials with properties similar to those obtained from irreplaceable resources or exhibiting harmful effects on the environment. In this respect, bioplastics have gained attention in quest of materials that can be used in place of conventional petro-chemical plastics. Biocompatibility, biodegradability and compostability of bioplastics are among the most favourable characteristics of the materials mostly derived from biological systems. Polyhydroxybutyrate (PHB) is a fully biodegradable bioplastic with similar physical properties to polyethylene and promising applications in various commercial fields including automation, aviation, medication, nutrition, fuel, packaging and many more. PHB production with Mixed Microbial Cultures (MMC) has recently gained attention as a cost effective production strategy by using bacteria that adapt with complex substrates presented in inexpensive waste materials. The initial research motivation was to enhance PHB production operation by means of the solutions obtained from sophisticated mathematical algorithms used for process optimisation. For this aim, a computer-based program simulating PHB batch process with MMC which was successfully validated with experimental data was available. Since mechanistic models of the simulation program could not be applied in optimisation algorithms, accurate empirical models were required. In the quest for reliable and accurate empirical models that can predict product concentration at the final stage of a batch operation, a methodology was developed in this study for classification of the batch operational regions based on the PHB critical process attributes. In the core of this research work, an innovative systematic methodology improves process understanding towards advanced process monitoring and control. This method enables operational scrutiny for generation of process knowledge regarding PHB process using MMC. The qualitative info-illustrations produced in the course of the classification method provide a sound platform for generation of considerably more accurate (quantitative) empirical models. These empirical models will be used in process optimisation studies. Abstract III In this research, PHB production occurs in a process type known as “feast and famine” or as “aerobic dynamic feeding” which is a well-known strategy applied for bacterial production with MMC. The “feast and famine” operations take place in Sequential Batch Reactors (SBR) in order to assure occurrence of the “feast” and the “famine” phases intermittently in each operational cycle. While PHB formation occurs during the “feast” phase, a “famine” phase should be followed to cause a cell physiological adaptation to maintain PHB production capability of bacteria. Establishment of the analytical methodology developed in direction of process empirical modelling realisation enables prediction of “feast” and “famine” phase occurrences based on the batch initial state documented for the first time in this work. This mathematical equation (“Phase Differentiating Equation”) plays a significant role in development of a novel SBR recipe for production of PHB with MMC. Execution of the recipe by the PHB process simulation program demonstrates high reliability of the proposed recipe. Application of the “Phase Differentiating Equation” in the SBR recipe assures favourable occurrence of the “feast” and “famine” phases in the majority of operational cycles. Reduction of operational failure rate reduces PHB production cost to improve its market position. The SBR recipe structure consists of six-stage cycles including (1) “feast” phase preparation stage, (2) “feast” phase operation, (3) operational quiescence, (4) product exploitation, (5) “famine” phase preparation stage and (6) “famine” phase operation. Operational reliability is investigated along with load disturbance rejection embedded in the SBR recipe. At the end, Sequential Quadratic Programming (SQP) is applied successfully as an optimisation algorithm to maximise PHB production under operational constrains.