One of the principal advantages of fuzzy rule based models over black-box approaches such as Neural Networks or polynomial models is transparency. The linguistic concept associated with the membership functions related to measured variables results in rules that are 'readable'. This quality is useful in analysing the functionality of processes through the model generated by data mining techniques. The greater the number of rules and the less descriptive the linguistic terms, the less transparent the model. The fewer rules, however, inevitably reduces the model precision with respect to the modelled process. This paper investigates the properties of Takagi-Sugeno models with either a linear function or singleton consequent with respect to model precision and transparency. The study is focused on a 'steady-state' heatexchanger model applied to the air-cooling process commonly found in heating, ventilating and air-conditioning (HVAC) equipment. The similarity measures are suitable to application to the on-line generation of these models.