Particle size measurement, specially in product streams, is fundamenta
l to control of a grinding circuit. In this paper, the design of a sof
t-sensor in an actual grinding plant is addressed, with the purpose of
increasing the availability of such measurement. The system is design
ed to substitute the momentary unavailability of the real sensor, eith
er because it has failed or it is in calibration or because it is time
shared by several parallel grinding lines. An ARMAX model structure i
s determined using the stepwise regression method. Starting with a lis
t of the presumably correlated candidate variables, the method selects
, one by one and in a systematic manner, the variables that best model
the measurement to be replaced. Not only the single measurements are
included in the list of candidates, but also combinations of such meas
urements having physical significance. A phenomenological model has be
en built with the purpose of determining these combinations. It turns
out that, in most of the cases, the components of combined measurement
s were selected by the method, instead of the single measurements. Act
ual grinding plant data is used to determine the model structure, to e
stimate the model parameters and to test the predictive capability of
the soft-sensors developed. Good results are obtained during periods o
f up to 24 h of the particle size soft-sensor operation. (C) 1998 Else
vier Science S.A. All rights reserved.