Ž . Chemometrics and Intelligent Laboratory Systems 50 2000 235–242 www.elsevier.comrlocaterchemometrics Monitoring of a wort fermentation process by means of a distributed expert system Alberto Bonastre a , Rafael Ors a , Miguel Peris b, ) a Department of System Engineering and Computers, Polytechnic UniÕersity of Valencia, Valencia 46071, Spain b Department of Chemistry, Polytechnic UniÕersity of Valencia, Valencia 46071, Spain Received 29 July 1999; accepted 19 October 1999 Abstract A distributed expert system has been proposed for the monitoring of a wort fermentation process through the flow injec- tion determination of total acidity, reducing sugars, ethanol and pH. Its configuration is mainly based on the use of dis- tributed nodes connected by means of a network and is capable of adapting itself to different situations. Satisfactory results have been obtained when it has been applied in a brewery plant, important advantages being shown over previously imple- mented centralized knowledge-based systems. q 2000 Elsevier Science B.V. All rights reserved. Keywords: Expert system; Process monitoring; Fermentation; Flow injection analysis; Distributed control systems 1. Introduction Expert systems are currently being applied to a greater or lesser extent in various fields of analytical wx chemistry 1 . The outstanding role they play in the monitoring and control of chemical processes under- going variations over the course of time, most no- tably fermentation processes, is especially remark- Ž . able. In this case, flow injection analysis FIA has proved to be of great value given its excellent perfor- wx mances and suitability to quality control analysis 2 . w x Peris et al. 3,4 have pioneered the application of expert systems to grape must fermentations, design- ) Corresponding author. Fax: q 34-96-387-7149. Ž . E-mail address: mperist@qim.upv.es M. Peris . ing a rule-based system that assists the on-line flow injection determination of some key chemical param- eters. It was a centralized system, whose scheme is presented in Fig. 1, implemented around a personal Ž. computer, which a has the necessary sensors and actuators to connect itself to the process to be con- Ž. trolled and b runs the knowledge-based system. The latter provides the inference engine, the rules, the Ž fuzzification and defuzzification operations numeri- cal values and logical values are converted into each . Ž other , and the timers which provide temporal rea- . soning . Nevertheless, this monolithic configuration shows some limitations. It usually lacks flexibility, in the sense that it is not able to grow or adapt itself to new requirements in an easy way. Additionally, the great deal of cables connecting the different parts of the 0169-7439r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. Ž . PII: S0169-7439 99 00065-9