Experimental testing for system backlash size relies on speed or displacement measurements of an isolated test apparatus but lack reference to engineered lash tolerances and expected values. These methods are prone to error when the measurement speed is not accounted for and hysteresis impacts the measurement. This study measures backlash on an experimental apparatus to update control model parameters. The results are also confirmed by comparing with analytical part tolerances based on CAE modeling. This is done by comparing three different test methods; output displacement only on a fixed-free apparatus, input torque with measured displacement on a fixed-free apparatus, and in-situ testing with torque and estimated displacement from rotating speed. The torque and displacement techniques recognize the influence of hysteresis and propose signal processing techniques to improve accuracy of the results. This signal processing technique was verified with an analytical model like those tested in the study. The three techniques used were all able to measure results within 31% of the CAE predictions when accounting for manufacturing tolerance. The signal processing method to account for hysteresis was analytically shown to have error less than 8.7% and 4.8% using an input frequency of <1 Hz with a sinusoidal and sawtooth forcing function respectively. The in-situ results were within 24% agreement of CAE projections without the need for a controlled laboratory test environment. These results showed that hysteresis should be accounted for when possible with an updated signal processing technique. The analytical model used to confirm backlash estimates with hysteresis also showed the need for controlled input forcing functions to improve the estimate accuracy. The study also confirms that experimental lash measurements can be collected from in-situ data when torque and displacement estimates are available.

系统间隙大小的实验测试依赖于隔离测试设备的速度或位移测量，但缺乏对工程间隙公差和预期值的参考。当不考虑测量速度并且滞后影响测量时，这些方法容易出错。本研究测量实验装置上的反冲以更新控制模型参数。通过与基于 CAE 建模的分析零件公差进行比较，结果也得到了证实。这是通过比较三种不同的测试方法来完成的；仅在固定自由装置上的输出位移，在固定自由装置上测量位移的输入扭矩，以及使用扭矩和旋转速度估计位移的原位测试。扭矩和位移技术认识到滞后的影响，并提出信号处理技术以提高结果的准确性。这种信号处理技术已通过研究中测试的分析模型进行了验证。考虑到制造公差时，所使用的三种技术都能够在 CAE 预测的 31% 内测量结果。分析表明，使用 <1 Hz 的输入频率，分别使用正弦和锯齿强制函数，考虑滞后的信号处理方法的误差小于 8.7% 和 4.8%。原位结果与 CAE 预测的一致性在 24% 以内，无需受控实验室测试环境。这些结果表明，在可能的情况下，应该使用更新的信号处理技术来考虑滞后现象。用于确认滞后估计的分析模型也表明需要受控输入强制函数来提高估计精度。该研究还证实，当扭矩和位移估计值可用时，可以从现场数据中收集实验间隙测量值。