Authors: B.M. Colosimo, F. Godio, L. Palmieri
Addresses: Dipartimento di Meccanica, Politecnico di Milano, Via Bonardi, 9–20133 Milano, Italy. ' Dipartimento di Meccanica, Politecnico di Milano, Via Bonardi, 9–20133 Milano, Italy. ' Dipartimento di Meccanica, Politecnico di Milano, Via Bonardi, 9–20133 Milano, Italy
Abstract: This paper originates from a real case study, in which the torque required for assembling automotive components has to be monitored. Since the data collected are truncated from below, this paper compares three different methods to design control charts for truncated data. The first method is the I-MR control chart. The second and third approaches proposed in this paper are based on estimating percentiles of truncated populations. In particular, the second approach is a non-parametric method, while the third one assumes a truncated normal distribution for the observed data. For all the methods, control limits are designed in Phase I, using the same false alarm rate. Performance of the three approaches are then evaluated by comparing the actual false alarm rate observed in Phase II to the one assumed in the design stage. The methods presented are finally rated in terms of the percentage error on the false alarm rate.
Keywords: SPC; statistical process control; control chart design; nonparametric control charts; truncated data; torque data; automotive components; component assembly; false alarm rate; automobile industry; quality control.
International Journal of Technology Management, 2007 Vol.37 No.1/2, pp.72 - 85
Published online: 23 Dec 2006 *Full-text access for editors Access for subscribers Purchase this article Comment on this article