Abstract The paper presents a systematic approach to the modelling and solution of a physical system using the decomposition based on what is called a process of homeostasy. A decomposed model of the physical system has been proposed, which instead of treating the system as a whole and solving a full system model, subdivides the model into submodels, which are solved using parallel processing. A decomposed model simulates the natural way the system behaves. In a real system a response to any disturbance is obtained simply by what can be called ‘exchange of information’ in points of connections of system elements. Each element acts independently, obeying its own submodel and acts in parallel with other elements. The mutual adjustment of the elements-states is achieved by mutual adaptation of their states while exchanging information with their neighbours. The method of adaptation is based on a weighted least-squares method. The final response of the system is obtained by an iteration process based on a process of homeostasis.