The paper presents a knowledge based approach applied to air pollution effects analysis in the case of PM2.5 air pollutant which has potential significant negative effects on human health. The use of knowledge derived from various sources (e.g. literature, databases, questionnaires, human experts' experience, and decision tables) via manual, semiautomatic and automatic methods is proposed for a multi-parameters analysis of the PM2.5 air pollution episodes effects on vulnerable people such as children and elderly. Some measures to reduce the negative effects on human health are also proposed by our approach. The knowledge under the form of production rules is incorporated in a knowledge base that is used by the ROKIDAIR intelligent decision support system (ROKIDAIR DSS). The knowledge base coherence was verified with the expert systems generator, VP-Expert. The experimental data sets that were used are for some air pollution monitoring sites situated in the Ploiesti city and included in the Romanian National Air Quality Monitoring Network.