The activated sludge system is the most widely used technology for biological wastewater treatment in the world. Its successful performance relies on the correct operation of both the bioreactor and the secondary settler. When settleability deteriorates, the inefficient separation of biomass can affect the quality of the activated sludge effluent, implying on most occasions an impact on the receiving ecosystem. These kind of undesired situations, known as activated sludge solids separation problems, are one of the main causes of inefficiency in activated sludge systems. They include: filamentous bulking, non-filamentous bulking, biological foaming, dispersed growth, pin-point floc and rising sludge. The origin of solids separation problems is (except from rising sludge) an imbalance between the different microbiological communities responsible for the biomass settleability: the floc-forming bacteria and the filamentous bacteria. Due to this microbiological origin, their identification and control is a tough task for plant operators. Knowledge-Based Decision Support Systems (KBDSS) are a group of tools from the Artificial Intelligence domain characterized by their capability to represent heuristic knowledge and to work with large amounts of data. The main objective of the present thesis is to develop and validate a KBDSS specially designed to support plant operators to handle solids separation problems of microbiological origin occurring in activated sludge systems. In order to achieve this objective, the developed KBDSS must accomplish with the following characteristics: (1) the implementation of the system must be viable and realistic in order to ensure its proper operation; (2) the reasoning process followed by the system must be dynamic and evolutive in order to match the necessities of the domain and (3) the reasoning must be also intelligent,. First of all, in order to guarantee the feasibility of the system, a thorough study, at a local scale (Catalonia) has contributed to the determination of the most common parameters generally used to diagnose, monitor and control these problems as well as to the detection of the existing limitations that the suggested KBDSS should overcome. The results obtained from past applications of KBDSS has demonstrated that the main bottleneck in developing KBDSS is the structure of the knowledge base (KB), where all the knowledge acquired from the domain is represented, together with the necessary reasoning processes. In our approach, the additional complexity and the corresponding necessities imposed by the dynamic nature of the domain exacerbate the limitations in developing a feasible system. In this case, a previous conceptualisation phase was considered in which a conceptual design of the system was set up. The Domino Model was suggested as a tool to conceptually design the system. Finally, in order to efficiently fulfill its main tasks, the last main objective or characteristic that the KBDSS must accomplish is the use of intelligent reasoning. In our approach, an Expert system (based on expert knowledge) and a Case-Based Reasoning System (based on experiential knowledge) have been and integrated as the main intelligent tools to carry out the goals of the KBDSS. In chapter 5 the development of the dynamic ES is presented. In chapter 6, a new temporal approach for classical CBRS is depicted, the Episode-Based Reasoning System (EBRS). Next, in chapter 7 some details of the KBDSS implementation in the G2 environment are presented. After that, in chapter 8 the results obtained during the 11 months of validation are depicted, including the results regarding the accuracy, adequacy, usefulness and usability of the system, which have been validated both experimentally (before the implementation) and as a result of the system’s implementation in the Girona WWTP. Finally, in chapter 9, the main conclusions derived from the present thesis are enumerated.