Application of indirect measurement techniques to an operating diesel engine

This paper deals with the indirect measurement of internal dynamic loads of an operating diesel engine. Direct measurement of excitations like piston slap or bearing loads is really difficult to implement with force sensors. Thus, their estimation is indirectly assessed from both a transfer model of the structure and operating response measurements. Techniques used in this work are limited to identification problems for which the studied machine is operating in stationary conditions, and all excitation areas are a priori known. There are two major difficulties when implementing such inverse approaches. On the one hand the transfer matrix inversion is known to generate an important sensitivity, i.e. a small change on input data (operating responses or transfer functions) causes a large deviation of output data (identified excitations). This difficulty is generally overcome by using regularization techniques as truncated SVD or Tikhonov. On the other hand, the reliability of the solution depends on the good concordance between the transfer model and the real operating structure. This second difficulty is a more penalizing one, because there is no mathematical solution to overcome it. In this contribution, internal excitations of the engine are identified from both acceleration measurements and two different transfer models. The first one is based on measurements and the second one on finite element computations. Truncated SVD is used to regularize the system. A pragmatic tool, known as the L-curve principle, is used to adjust the regularizing parameter. Forces identified with either measured or computed transfer functions are compared between one another to assess their reliability. A time-domain recomposition method is introduced, and time-domain identified excitations are shown. Finally, internal sources are ranked in function of their contribution to the timing belt side mount acceleration.