主催: The Japan Society of Mechanical Engineers
会議名: 動力・運動伝達系国際会議MPT2017
開催日: 2017/02/28 - 2017/03/03
We proposed a new signal processing method to estimate angular position of gears without help of speed sensors. In the method, a neural oscillator, which was able to synchronize with meshing vibration measured by an accelerometer, was used, and the angular position of gears was estimated from the period of the synchronized oscillator. The design method of oscillator's parameters, such as time constants, coupling coefficients, etc., was shown to obtain desired amplitude and frequency. However, there was no way to determine the input gain of the oscillator. Generally, synchronization region on a frequency axis broadens with an increase in the input gains of oscillators, and length of time it takes to synchronize is saved with the increase. In other words, saving the length of convergence time by an increase in the input gain worsens the susceptibility to noise. Therefore, the input gain of neural oscillator should be determined with consideration for the design trade-off. The long-term objective of this study is to show the design method of the oscillator's input gain. As a very first step, this paper shows that the estimation error of angular position of gears is affected by the input gain change. Especially, we show the effect, that several different neural oscillators have, on the estimation errors with an increase in the input gains. The target neural oscillators have natural frequencies, which are set at the meshing frequency of a gear pair of interest or not. Acceleration responses were measured with a pickup set at the top of a housing of a driving-side gear shaft bearing during gear operation tests, and input to the neural oscillators. Then, synchronous generation and the estimation error of the proposed method had were investigated. As a result of this investigation, it was concluded that the estimation error in the case of the neural oscillators, whose natural frequencies were shifted from the meshing frequency, have the local minimal values with an increase in the input gains, useful information for determine the input gain of the neural oscillator was included.