Tool condition diagnostics system is still a missing element of advanced cutting machines. Tool condition monitoring is an important problem in the furniture industry and related industries. An effective diagnostics system would increase machining efficiency, while increasing machining accuracy and tool reliability. It would also allow for flexible automation of the manufacturing process. This paper presents a method for estimating wear of tool blades based on multiple measures of diagnostic signals. A typical solution to the problem of approximating blade wear based on multiple measures is the use of a multilayer neural network of feed-forward and back-propagation algorithm. The process of optimizing the architecture and parameters of the neural network is presented.