Patents are an increasingly important source of technological intelligence that companies can use to gain strategic advantage. They can be used as a stimulus for R&D to search for whether someone somewhere has already solved the problem or a very similar one. In this way, the answer to technical questions depends on how we are able to extract crucial information from the patent corpus and translate it into knowledge. The state of the art of IR tools for patent searches is very rich and in continuous improvement; moreover, current tools are inadequate to satisfy users’ expectations. This paper gives a general overview of the universal tools for knowledge management and proposes a combination of knowledge bases, design methods like TRIZ and FBS theory and physical effects for improving function-based patent searches. An example dealing with a new design of nutcracker proposes the use of keywords related to physical effects in order to search a non-clearly expressed function (or behaviour).