This paper proposes a new way to estimate the flow in amicromechanical flow channel. A neural network is used to estimate thedelay of random temperature fluctuations induced in a fluid. The designand implementation of a hardware efficient neural flow estimator isdescribed. The system is implemented using switched-current techniqueand is capable of estimating flow in the μl/s range. The neuralestimator is built around a multiplierless neural network, containing 96synaptic weights which are updated using the LMS1-algorithm.An experimental chip has been designed that operates at 5 V with a totalcurrent consumption of 2 mA, resulting in a power consumption of 10 mW.The dimensions of the clip core are 3 mm×4.5 mm
展开▼