Rotational Speed And Vibration Measurement Through Digital Imaging And Image Processing

Rotating machinery plays an important role in energy,chemical,power and metallurgical industries.The rotational speed and vibration of rotating machinery are important operating parameters,which can be used to obtain the health status of rotating machinery timely and accurately.Meanwhile,they can provide scientific data for early warning of system failure and unplanned shutdowns to improve operation safety and efficiency of mechanical devices.However,existing measurement systems for such purposes have their limitations in terms of measurement performance,cost and scope of application,especially in complex and hostile environments.For example,conventional contact-type measuring devices installed on micro electro mechanical systems can introduce additional mass,which will affect the dynamic characteristics of mechanical systems.For some measured objects,such as large wind turbine blades,the installation and maintenance cost of conventional sensors is high.Therefore,better instruments should be developed using new principles and methods to achieve accurate and reliable speed and vibration measurement of rotating machinery.With rapid advances in imaging sensors and image processing algorithms in recent years,imaging-based techniques provide a promising alternative to rotational speed and vibration measurement.However,existing vision measurement systems mainly use expensive high-speed cameras to acquire clear images.The low affordability of high-speed cameras limits the wide application of such systems.In addition,due to the influence of motion blur,the measurement accuracy of existing approaches needs to be improved.A measurement system based on a low-cost image sensor and image processing for speed and vibration measurement is proposed in this reserach through studying the motion characteristics of rotors and the induction mechanism of image sensors.Compared with measurement systems with high-speed cameras,the proposed measurement system can greatly reduce measurement cost.The main research contributions are as follows:(1)A novel method based on image similarity level is proposed to measure rotational speed.A low-cost image sensor is used to acquire an image sequence.Image similarity evaluation algorithms are used to convert the image sequence into a one-dimensional signal indicating image similarity level.Meanwhile,new instantaneous frequency estimations methods,including short-time Chirp-Z transformand and short-time autocorrelation based on parabolic interpolation are proposed to process the reconstructed signal of similarity level and extract rotational speed information.The validity of the method is verified through theoretical analysis and experimental tests.Experimental results demonstrate that the proposed system can provide measurements of a constant or gradually varying speed with a relative error less than±0.7%over a speed range from 100 to 3000 revolutions per minute(rpm).Under variable speed conditions,this system is capable of providing valid speed measurement with a maximum error of 1.4%.(2)A novel method based on image correlation is proposed to measure rotor vibration.When a rotor is in motion,the grayscale distribution of rotor images varies with the vibration of the rotor.Image correlation algorithm is used to process the image sequence of a rotating rotor to obtain an image correlation signal.Finally,the vibration frequency and relative vibration amplitude can be obtained from the spectral analysis of the image correlation signal.Experimental assessments of the proposed method were conducted on a purpose-built test rig.A commercial eddy current sensor was used to provide reference data.Experimental results demonstrate that the vibration frequencies and their relative amplitudes obtained from the proposed measurement system agree well with those from the eddy current sensor.Meanwhile,the rotor vibrates at several distinct frequencies that increase with the rotational speed,which agrees the expected vibration characteristics of the rotor.(3)A novel method based on a QR marker tracking is proposed for rotational speed measurement.Local feature extraction and image matching are used to detect the QR marker attached on a rotor and thus measure the rotating angle and speed of the rotor.In this paper,AKAZE local feature is adopted to achieve object dectection.To improve the accuracy of feature matching accuracy and reduce feature mismatching,image correlation coefficient and Hamming distance is used to optimize feature matching process.Experimental results show that the maximum relative error of speed measurement is less than ±1%,and the normalized standard deviation is less than 0.6%.The proposed method performs better than tradtional SIFT and SURF algorithms.This method only needs two rotor images to achieve rotational speed measurement with high measurement accuracy and good real-time performance.(4)The performance assessment of the proposed measurement system is conducted on a rotor test rig under a variety of experimental conditions.The experimental results verify the reliability and environmental adaptability of the measurement system.The effects of image resolution.frame rate,exposure time,shooting angle,light condition on the performance of the measurement system are studied to determine the best operational parameters of the visual measurement system.Field tests of the proposed vision measurement system,including the rotational speed and vibration measurement of wind turbines and the speed measurement of a motor shaft,are carried out.Experimental results show that the developed measurement system can measure the motion information of rotors accurately,such as the rotating speed and vibration.Compared with an existing imaging method,the proposed imaging approach performs better in terms of measurement accuracy and speed range.