Abstract While moving towards complete automation of the grinding process in order to be able to realise unattended manufacturing, it becomes mandatory to closely monitor the process in order to detect any malfunction at the earliest moment with high reliability. In the grinding process, a proper estimate of the life of the grinding wheel is very useful. When this life expires, redressing is necessary. Generally, chatter marks, surface roughness, burn marks, etc. are considered as the tool-life limit in grinding. In this paper, the occurrence of burn marks on the work surface is adopted as a criterion of the wheel life; accordingly, the time of the occurrence of grinding burn during the cylindrical plunge grinding of steel is studied under different conditions of grinding. The data thus collected is used for the prediction of the time to burn using an artificial neural network.