KEYWORDS: Point clouds, Image registration, Cameras, 3D image processing, Matrices, 3D acquisition, Speckle, Imaging systems, Feature extraction, Singular value decomposition
In response to the repetitive surface features or symmetrical structures of most parts in the industrial field, with smooth and texture-deficient surfaces, traditional point cloud registration algorithms struggle to extract correct three-dimensional correspondences, leading to misregistration and getting stuck in local optima. This paper proposes a point cloud registration method that utilizes texture image pre-registration and multi-criteria mismatch suppression. Firstly, speckle images are projected onto the object under test, obtaining point cloud and image data from different perspectives.Then, the optimal registration region is located in the grayscale image, and feature points are detected and registered in this region, obtaining corresponding three-dimensional feature points based on the coordinate transformation relationship. Simultaneously, a multi-criteria mismatch suppression method is applied to eliminate mismatched point pairs. Through Singular Value Decomposition (SVD), achieving coarse point cloud registration. Finally, a pose optimization is performed using the Iterative Closest Point (ICP) algorithm on the pre-extracted overlapping regions. Experimental results demonstrate that this method provides efficient and accurate registration for point clouds of parts with surfaces lacking texture and featuring repeated patterns.
The effectiveness of the Complex Steerable Pyramid (CSP) in enhancing subtle structural motions for dynamic facility observation is widely acknowledged. However, further analysis and evaluation are needed to determine the applicability of directly utilizing CSP for displacement measurement in engineering applications. This paper assesses the displacement measurement performance of phase-based optical flow (PBOF) using HalfOctave CSP. Initially, the impact of the CSP's down-sampling rates on the accuracy of displacement measurements in PBOF is assessed by simulating motion to identify suitable configurations. Afterwards, the accuracy of this method is investigated within the effective range of displacement measurement by simulating motion. The results of the simulation experiments indicate that the optimal down-sampling factor for CSP should be set to zero. When CSP is at its optimal down-sampling factor and using filters located in the middle of the effective filter level, PBOF achieves the best comprehensive displacement measurement performance.
KEYWORDS: Data modeling, Wind turbine technology, Frequency converters, Instrument modeling, Data conversion, Statistical modeling, Fluctuations and noise, Analytical research, Wind energy, Temperature metrology
In recent years, with the development of wind power industry, the installed capacity is increasing rapidly, and the installation sites are developing towards the ocean and remote mountainous areas, which makes the maintenance of wind turbines difficult and the cost increases. In order to reduce the maintenance cost and improve the maintenance efficiency, and ensure the economic and reliable operation of the equipment, many state recognition and fault early warning methods have been proposed one after another, and the intelligent fault diagnosis method based on supervisory control and data acquisition(SCADA) data and various kinds of machine learning has gradually become a research hotspot. In this paper, the generator system of a wind turbine in a wind farm is taken as the research object. By using the massive SCADA data recorded in operation, a multivariate state estimation technique (MSET) is established to predict specific operation parameters with relevant operation parameters as input. This method uses clustering algorithm to clean up the selected SCADA data, then uses MSET to establish the prediction model, and calculates the prediction residual by sliding window method to realize the fault diagnosis. Finally, the effectiveness of the method is verified by actual SCADA data
Nowadays, the number of assembled wind turbines in the world is growing more rapidly, which brings an urgent need for intelligent operation and maintenance of wind turbines. The intelligence of wind turbine operation and maintenance is based on the high-precision classification and recognition of SCADA system data. In response to this demand, this paper establishes a wind turbine normal data discrimination model that combines SCADA system data preprocessing and random forest integrated learner. First, obtain a determinable sample dataset according to the principles of statistics and the NearMiss under-processing method. Then build a decision tree, use the features in a variety of SCADA datasets to train and learn the sample dataset, and form a random forest to determine the normal data model of wind turbines. The results show that the model can effectively classify whether the SCADA data of wind turbines is normal, achieve a higher accuracy rate, and improve the reliability of discrimination, which is of great significance to the subsequent research on intelligent operation and maintenance of wind turbines.
There are many redundant parameters in Convolutional Neural Network(CNN) when it is used for target recognition in a specific scene, which will greatly occupy the calculation amount and affect the operating efficiency of software and hardware, and cannot meet the real-time target detection requirements of the algorithm in a specific scene. In this paper, channel pruning, layer pruning and their hybrid pruning experiments were carried out on you only look once version 3(YOLOv3), a typical CNN target recognition model. The result of the hybrid pruning can greatly reduce the model parameters and the amount of calculation, and can reduce the model of resource utilization through the comparative analysis. And the volume of the model after hybrid pruning was reduced 94.4% when mAP only loss 0.9%. The model inference time was reduced by 36.6%. This study could provide references for the optimization of object recognition model in structured scenes such as road and workshop.
The production process of robot castings mainly includes processes such as robot core taking, core assembly, core setting, dipping, pouring, sand casting, polishing, and cleaning. The handling process involves every process of the casting production line and is one of the key processes of the casting production line. For the 6-axis robot used for handling, the kinematics model of the robot is carried out according to the improved DH parameter method, and the dynamic model of the robot is established by using Newton-Euler equation and Lagrangian equation respectively. Secondly, according to the requirements of casting handling technology, the motion trajectory is planned according to the established kinematics model and dynamics model, and the dynamic simulation is carried out with MATLAB. The joint torques of the robot dynamic models established by the two methods are very close, which verifies the correctness of the dynamic model.
In this paper, the converter of a 2MW direct-driven permanent-magnet wind turbine is researched for exactly locating the open-circuit fault of wind power converter. The structure and fault characteristics of the converter is analysed, and open-circuit fault location of the converter is implemented by defining a fault feature variable based on grid-side current signal. Based on Matlab/Simulink platform, the fault simulation model of wind power converter is built, the waveforms changes of grid-side current and fault feature variables are analysed under different fault conditions. The simulation results show the feasibility and effectiveness of the model. It provides a fast-position diagnosis method for the open-circuit fault of wind power converter, and has the certain application value for project.
The working state of the pitch motor has a great influence on the operation of the wind turbine. In this paper, the 2MW wind turbine is taken as the research object. Based on the historical operation data of the wind turbine, the influencing factors of the pitch motor temperature deviating from the normal range are analyzed. First, the range of the pitch motor temperature is counted by the quartile method. Then, using Relief-F for feature selection, the characteristic parameters that have a great influence on the pitch motor temperature are screened out. According to the selected characteristic parameters, combined with the historical operation data of the wind turbine, the influencing factors of the abnormal temperature of the pitch motor are analyzed. Through analysis, it is found that the blade pitch angle, the pitch motor current and the battery box temperature show obvious trend changes in the period before and after the abnormality of the pitch motor temperature. They are related factors that affect the abnormal of the pitch motor temperature. The main meaning of this paper is to screen the characteristic parameters that affect the pitch motor temperature through the Relief-F algorithm. The selected characteristic parameters are representative and can be used as the input parameters of the prediction model of the pitch motor temperature. It can better deal with the difficult problem of parameter selection for early warning model modeling.
Many researchers have done a lot of research in the fields of welding automation and welding robots. The existing technology can achieve better results in applications where the shape and position of the weld are relatively fixed and the weld is easier to identify. For welds with inconspicuous appearance, irregular shape and complicated background such as the casing weld on copper pipe of refrigerator compressor. The existing technology is difficult to achieve stable and accurate automatic positioning and tracking of welds. In this paper, a three-dimensional positioning method for casing welds of copper pipe based on binocular vision is proposed. Firstly, Zhang's calibration method is used to compute the intrinsic and external parameters of the binocular vision measurement system. Secondly combine with morphological operation and basic image operation, the casing welds in complex scenes are segmented accurately. Finally, the binocular measurement model is used to complete the three-dimensional positioning and tracking of the casing welds in complex scene. The measurement results show that the measurement accuracy of the system can meet the requirements of vision guided automatic welding.
Target tracking is an important research direction in the field of machine vision. How to quickly track the target position in the time-series images is a key issue that has been widely studied. The search method determines whether the best motion vector can be quickly found, it also has an important impact on the efficiency of the measurement. The paper proposes a fast template search method for multi-scale moving target tracking, which can achieve tracking and searching targets in dynamic scenes, and measures the moving distance of the target. Compared with full search algorithm and three-step search algorithm, the measurement method in this paper can reduce the number of searches on the basis of guaranteeing the measurement accuracy. The search algorithm can avoid trapping in local optimum. For the content of the research, a discrete sampling method is proposed to solve the problem of starting point and initial step setting of the three-step search algorithm.
KEYWORDS: Wind turbine technology, Data modeling, Failure analysis, Temperature metrology, Standards development, Wind energy, Data acquisition, Analytical research
The aim of calculating the standard value of the stator iron-core temperature of the wind turbine is to predict the wind turbine fault. At first, Simplify the calculation model of turbo generator stator iron-core temperature which temperature rise mechanism similar to that of wind turbine. The model uses the generator active power, reactive power, voltage and other operating parameters to calculate the generator stator iron-core temperature. Second, the parameters in the model were identified by the first year’s normal rated state operation data of the wind turbine collected by the wind farm SCADA system. Then, the model obtained after identification can calculate the standard value of the stator iron-core temperature under the rated working state of the wind turbine. Finally, If the standard value calculated by the model is correct, the average difference between the standard value and the measured value is related to the wind turbine failure alarm information. After calculate the standard value of the stator iron-core temperature using the fifth year’s rated state operation data, it’s easy to see that when the average difference between the standard value and the measured value is large, the wind turbine generates an alarm fault message in the adjacent time period and when the average difference between the standard value and the measured value is small, the wind turbine runs normally in the adjacent time period. This correlation indicates that the simplified model can correctly calculate the stator iron-core temperature of the wind turbine.
Structural crack is an important factor which causes failure of reinforced concrete bridges. In this work, automatic detection and dimensional measurement of concrete bridge crack are researched, for improving technical level and efficiency of concrete bridge state assessment. Images containing crack features are first recognized using information entropy characteristics of intensity clustering, for promoting efficiency and robustness of rough crack localization based on proportional segmentation. After the features are refined at sub-pixel level, their actual dimensions are accurately measured employing a cross structured light system. Experiments show that the problems such as high misjudgment, low efficiency and poor accuracy in the existing technologies are preliminarily addressed; the proposed method performs well in crack detection and measurement using concrete bridge structure images.
In order to achieve automatic welding of casing welds in complex industrial environments, a casing weld visual identity method is proposed. Firstly, the pipe is detected by combining grayscale characteristics and color characteristics. Secondly, the weld potential regions are obtained by the morphological operation. Finally, weld seam identification is realized based on the positional characteristics of the casing weld on the copper pipe. Experimental studies verified the effectiveness of the method in complex environments.
It's necessary for automobile to detect and adjust four-wheel alignment parameters regularly, due to the significant effect on improving stability, enhancing security and reducing tire wear of automobiles. In order to measure the parameters that determined by relative position and posture of four wheels to the automobile cab, this paper proposes a method which applies monocular vision of linear structure light to wheel pose measurement. Firstly, space coordinates of feature point cloud are calculated out from the principle of structured light. Then, an algorithm is designed to determine the normal vector of wheel tangent plane and measure the wheel pose. Finally, actual experiments that by evaluation of adjusted wheel angle measurement are carried out to verify the system accuracy. The corresponding studies can be applied in designing and developing 3D four-wheel alignment system that based on structured light.
In view of the faults of traditional method, a computer vision-based swing center testing method for flexible joint was presented. The first step is to obtain original measurement date by a single industrial camera continuously capturing images of two circular marks which were fixed in the swinging object. The second step is to achieve stable and accurate multi-feature extraction in complex environment by a circular feature extraction method presented by our research group. And the final step is to obtain test results through the markers center coordinate and the swing center computing model. The experimental results indicate that the peak value (PV) error of this method is 1.6mm. This method is effective and it is better than traditional testing methods.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.