Exponential smoothing of learning curve data
DENIS R. TOWILL · International Journal of Production Research · 1977
SUMMARY The paper investigates the effect of exponentially smoothing learning curve data plotted in the time constant model form. In the time domain, the result is a smoothed, delayed version, of the raw data. Attenuation of scatter about the trend line is predicted using the z transform method to establish the exponential smoother frequency response. The optimum time shift for trend correction is determined by an analogue method based on the Matsuburu time delay theorem, the result being identical to that obtained using the ‘ equivalent age of data ’ concept. Results for six industrial operators performing light assembly tasks confirm the optimum time delay prediction as judged by the minimization of mean square error criterion, and the frequency response analysis is confirmed on application to a learning curve describing crew performance during a simulated naval underwater mission.