Developing an expert system for diagnosis of problem gas chromatographic data

Abstract An expert system with a user-friendly interface has been developed to aid in the diagnosis of faulty analysis by gas chromatography as a first step towards totally automated analysis. The system advises users on the identity of faulty components and the possibility of improper operation based on the appearance of the chromatograms that result from the experiment. The knowledge used in this expert system has been assembled into a knowledge domain matrix and was coded using case histories. One case serves as an example: “A leaking syringe will result in loss of signal intensity and the appearance of false peaks due to contamination”, is coded to give two entries in the knowledge domain matrix the cause “leaking syringe” is “True” for both “sensitivity loss” and “syringe related peaks”. The knowledge base can be readily expanded through use of Table Edit and Rule Edit (programs developed in our laboratory). The inference engine (EAengine) can apply forward, backward or mixed chaining strategies to obtain a cause for the symptoms that are observed. The system is written in Microsoft Visual BASIC 2.0 within Window 3.x, calling the inference engine as a dynamic link library (DLL). The user can enter observations easily using a powerful new graphical-based interface. Gas chromatograph operation and control systems that use Windows 3.1 may be in progress at the same time. The diagnostic program may be suspended while a more urgent task is run. Communication between a Varian Star Workstation and this system was realized through the dynamic data exchange (DDE) facilities of the Windows 3.1 shell.