High-speed camera monitors robot vibrations without needing sensors
03/03/2026
Japanese researchers have devised a non-contact system for monitoring the vibrations of industrial robots as they operate, without using costly physical sensors that can be affected by interference. The researchers, from Hiroshima University, are using high-speed cameras and a non-contact optical technique called digital image correlation (DIC) to measure surface deformation, displacement and strain in the robots. The technique could be used for predictive maintenance.
Abnormal vibrations in industrial robots can signal mechanical issues such as imbalances, loose components, worn bearings or structural weaknesses. These issues can lead to unexpected shutdowns, costing thousands of pounds per hour in lost production. Poor maintenance can also lead to unpredictable robot behaviour, creating hazards for workers.
Physical sensors such as piezoelectric accelerometers can be highly effective for monitoring vibrations in robots, but they have drawbacks, particularly their high cost and the fact that they add weight to moving robotic arms. Wired sensors also need cables that trail along the arms, potentially resulting in problems such as flexing, wear, tangling or breakage.
The researchers used a six-degrees-of-freedom (6DOF) vertical articulated robot as their measurement target. After applying a random, high-contrast speckle pattern to the robot, they attached a small vibration exciter.
They captured video footage of the moving robot using a Mikrotron EoSens 2.0CXP2 camera with a 20 mm focal-length lens, located 1.2 m from the robot.
The video footage was divided into individual frames based on a frame rate of 1000 frames per second. Each 1920 × 1080 pixel frame was divided into 128 × 128 pixel blocks, yielding 435 sub-images that simulated vibration sensors.
A vibration visualisation algorithm tracked changes in the speckle pattern on the robot, generating detailed 2D maps of how the surface moved as a result of vibrations.
The system captured simultaneous measurements across the robot, detecting both high- and low-frequency vibrations. It achieved a level of accuracy that the researchers claim would be difficult to reach using sensor-based techniques.
The scientists now plan to enhance the robustness and practical applicability of their technique. They want to extend it to perform real-time vibration monitoring while the robot is moving, enabling more accurate characterisation of its operational dynamics. Other robot configurations will also be tested and the researchers may turn to multi-camera stereo imaging to enable full 3D vibration analysis.
