ENVOPExpert: An Intelligent Tool for Automating 
Waste Minimization Analysis in Chemical Plants

The issue of pollution prevention has evolved significantly within the chemical industries over the past decade. The end-of-pipe approach, which has been traditionally the chief solution to pollution prevention, is no longer viewed as adequate nowadays. Increasing public awareness, escalating waste treatment and disposal costs, and more stringent discharge standards have put enormous pressure on chemical industries to find better solutions. In Singapore, the problem is further complicated due to the scarcity of landfill space for final waste disposal. The concept of waste avoidance or minimisation holds much promise for the chemical industries in Singapore in the coming years.

Waste minimisation is defined as any technique, process or activity, which prevents, eliminates or reduces waste at its source, or allows for reuse and recycling of waste. Several different waste minimization techniques have been previously described in literature. These include pinch technology, numerical optimization and ENVironmental OPerability (ENVOP) study. ENVOP study is a technique that follows the approach of HAZard and OPerability (HAZOP) analysis commonly used in industry for process safety analysis. Waste minimization is a team-based activity, in general. During a brainstorming session, the team assesses various different parts of the process of concern to generate several ideas for reducing waste. All of these are then followed up for identifying the most feasible option - usually on the basis of technical and economical feasibility. A thorough waste minimization programme requires specialized expertise and is laborious, time-consuming, expensive and knowledge-intensive. This has caused a major technical barrier for implementing waste minimization within the industry. An intelligent system that can automate waste minimization analysis would certainly be beneficial since it can perform a systematic and thorough evaluation of waste minimization options and reduce the team's time and effort.

Research on developing an intelligence tool for automating waste minimization analysis is ongoing in the Laboratory of Intelligence Applications in Chemical Engineering (iACE) of the National University of Singapore. An intelligent, general-purpose expert system called ENVOPExpert has been developed as part of the research project. The scope of ENVOPExpert can be defined as follows: Given specific information about a chemical plant including the process flowsheet and process chemistry comprising of the reaction and separation schemes, the system automatically identifies all waste causing activities and recommends ways to minimise them. ENVOPExpert performs these tasks using generic waste minimisation domain expertise that has been developed and implemented in its inference engine. The domain expertise comprises various heuristic rules, procedures and methods for common process units including reactors, distillation columns, heat exchangers, compressors, etc. These domain knowledge parameters are automatically integrated with user specified process information to perform waste diagnostics and identify various waste minimisation alternatives.

ENVOPExpert has been implemented using Gensym corp's G2 expert system shell. The user interacts with ENVOPExpert using a graphical user interface, as shown in Figure 1. The results of the waste minimisation study are presented as a table of various waste minimisation alternatives that can be further explored. These results can also be grouped in several ways such as all the wastage in a particular process unit, or all the wasteful expenditure originating from each material, etc.

Figure 1: A portion of ENVOPExpert's graphical user interface.

We have successfully tested ENVOPExpert on an industrial case study involving a hydrocarbon separation process. In general, the system is able to derive the same waste diagnostics and solutions as those concluded by the team of experts. While the results are very encouraging, currently we are testing the system on a larger and more complex industrial case study. We are also adding additional capabilities to ENVOPExpert. One of these is the ability to automatically rank the waste minimisation alternatives on the basis of process economics. We are also seeking potential users from the industrial community for testing and using ENVOPExpert on real-life case studies.