Testing and Validation of Measurement Methods for Reducers and Gearboxes

　　In the measurement and control system, the motor control module sends commands to the motor driver, which in turn applies the corresponding torque to the motor, enabling the shaft system to transmit the specified torque. Simultaneously, the torque sensors in both the instrument and the torque calibrator experience the same torque. During this loading process, the frequency signals output by the torque sensors in the instrument and the torque calibrator are routed to the data acquisition module, which captures these signals and validates them as the raw measurement values. The signals acquired by the torque calibrator serve as the standard reference for the input and output torques of the robot’s reducer. By comparing the synchronously acquired reference values with the raw measurements, the errors incurred during calibration can be determined. The data processing module then uses these error values to fit a compensation curve, implemented via a BP neural network. During the measurement of the reducer under test, this compensation model is applied to correct the torque signal errors, and the corrected torque data are finally output and displayed as the final measurement result.

　　 Testing and Validation of Measurement Methods

　　Following the methodology described in the design and taking into account the structural characteristics of the vertical‑type instrument, a torque calibration system was established. In the torque sensor calibration experiments, the torque calibrator was positioned at the test station, and the calibration procedures were carried out according to the steps outlined in Section 4. The motor within the instrument applied torque incrementally in both clockwise and counterclockwise directions, with the output torque being recorded. A BP neural network was then employed to compensate for deviations across the torque range, achieving the specified measurement accuracy. The error curves were generated using the method described in Section 4.

　　 Error fitting curve of TMMOSR output torque

　　The compensation performance of the instrument’s additional torque is verified using a higher‑accuracy torque standard accredited by the National Institute of Metrology, China, thereby determining the torque output accuracy of both the TMMISR and the TMMOSR. The TMMISR has a torque measurement range of ±50 Nm, while the TMMOSR covers ±2000 Nm; the positive or negative sign indicates whether the applied force is clockwise or counterclockwise. Within each range, 50 sampling points are selected for comparative analysis. Figure 11 presents a set of results comparing the TMMOSR’s output torque with the standard torque, before and after error compensation.

　　 Error of the torque measurement system before and after compensation

　　The experimental results indicate that the maximum error is 1.8 Nm, corresponding to 0.09% of the full measurement range. Through ten validation runs—incorporating repeated measurements and the selection of additional sampling points—the instrument achieves a torque measurement accuracy of up to 0.1% of full scale.

　　 Conclusion

　　This paper presents a torque measurement system designed to measure the input and output torques of high-precision reducers in industrial robots. The torque sensor is positioned close to the robot’s reducer to minimize the measurement path length. Structural optimization has been employed to enhance the overall stiffness of the system.

　　The symmetrical structure, featuring a cylindrical measurement module and a disc‑type support, simplifies the deformation model and facilitates error compensation. By identifying the primary factors influencing torque measurement errors and analyzing their two main components, the characteristics of these errors are elucidated. To enable calibration and error compensation, a torque calibrator was designed to calibrate the torque sensor within the torque measurement system. A BP neural network was employed to compensate for torque measurement errors. Experimental results validating the effectiveness of the error‑compensation model demonstrate that the precision of the torque measurement system reaches 0.1% of full scale.

　　Furthermore, most torque measurement errors have been compensated. All design and calibration methods ensure that this instrument delivers superior performance compared to other commonly used testing devices.