To design an instrument to measure numbers per ml of particles in the range 1 to 25 microns diameter, in a 35 mm diameter bore carrying fluid moving at up to 6 m/s, at 400 bar and 90 ºC, over a range of cleanliness typical of hydraulic fluid applications, without causing the bias typical of a small bore side-flow, and at low cost of components.
Optical components were used to provide a structured light field in the bore with which particles interact to give ambiguous signals. Bayesian information location and Bayesian inference were used to analyse a large number of such ambiguous interactions and hence deduce the particle size distribution after sufficiently many interactions have been observed. We implemented this algorithm on a low-cost DSP chip from which information to build a graphical result was exported to a controlling PC.
By allowing the initial construction of ambiguous rather than unambiguous signals hardware cost was minimised and one-off development cost focussed in software. By correctly managing uncertainty throughout the algorithm, the uncertainty in the resulting distribution is available to the user at any time based on the interactions observed to date. By keeping the hardware simple it was possible to meet the demanding environmental specifications and meet the challenging requirements on bill of materials.