PREDICTION OF TOOL FAILURE FROM A PROBABILISTIC POINT-OF-VIEW

Tool performance and tool life are important criteria in the designing and optimizing of forming processes. Tool behaviour and service life can be analysed by computer aided modelling, taking the whole process into consideration. However, the result of such a process simulation i s usually a single value of life time, not reflecting the high degree of its uncertainty that characterizes the life time, as is well known from practice. Thus, the necessity of taking into account the stochast ic characteristics of tool life is quite evident. The objective of fai lure analysis should be the prediction of life time for a well defined level of confidence. A systematic approach to this problem reveals th at the evolution of failure probability can be derived by quantifying the interaction of load on the tool and tool strength, both being infl uenced by deterministic as well as by stochastic factors. Hence, a pro mising way to come to a reliable failure prediction is given by the co mbination of mechanical/numerical methods and statistical analysis, i. e. by the integration of statistics into process simulation. The benef it of such an approach is illustrated by two examples in which statist ical methods are applied to achieve an improvement of tool performance . In the first example the reduction of strength dispersion is shown t o be very effective in decreasing the failure probability. Provided th at the stochastic character of both load and strength can be quantifie d, the most promising measures to reduce strength dispersion can be fo und by process simulation. In the second example the shape optimizatio n of a cold-forging die is considered. The effect of optimization can only be assessed if the topographical features of the tool surface are taken into account, which can be realized by applying a combination o f fracture mechanics and statistical concepts. The result is again a q uantifiable probability of failure.