I think every software-based system has a Generic Software Quality Evaluation mechanism which could be functional and or structural. However, unlike any other software system, performance of any knowledge processing system is equally dependent on the Knowledge representation and the inference algorithm.

Hello Stefan, when you want to validate quality of knowledge base (core of knowledge or expert system) you can find some articles concerning benchmarking of expert systems. When you want to make expert system with professional software, it has control mechanism for knowledge acquisition process (if you have some incompability etc.) See the pages: http://www.exsys.com/, http://xpertrule.com/ etc.

The process of creating appropriate knowledge base is not easy. For example, when we want to create adequate rule-based fuzzy system from surveys, we need to use more then 11000 rules and the question is: how to optimize a lot of rules ;) 

The design of ES depend: on the knowledge type, on the character of problem, on inferencing and strategies etc.  To design ES is more art then engineering ;)  

I do not know it is exact what you mean, but good luck in your research. 

http://www.exsys.com/

http://xpertrule.com/