Parts with high-quality freeform surfaces have been widely used in industries, which require strict quality control during the manufacturing process. Among all the industrial inspection methods, contact measurement with coordinate measuring machines or computer numerical control machine tool is a fundamental technique due to its high accuracy, robustness, and universality. In this paper, the existing research in the contact measurement field is systematically reviewed. First, different configurations of the measuring machines are introduced in detail, which may have influence on the corresponding sampling and inspection path generation criteria. Then, the entire inspection pipeline is divided into two stages, namely the pre-inspection and post-inspection stages. The typical methods of each sub-stage are systematically overviewed and classified, including sampling, accessibility analysis, inspection path generation, probe tip radius compensation, surface reconstruction, and uncertainty analysis. Apart from those classical research, the applications of the emerging deep learning technique in some specific tasks of measurement are introduced. Furthermore, some potential and promising trends are provided for future investigation.
6 March 2018 in Measurement Science and Technology
Circular targets are commonly used in vision applications for its detection accuracy and robustness. The eccentricity error of the circular target caused by perspective projection is one of the main factors of measurement error which needs to be compensated in high-accuracy measurement. In this study, the impact of the lens distortion on the eccentricity error is comprehensively investigated. The traditional eccentricity error turns to a quasi-eccentricity error in the non-linear camera model. The quasi-eccentricity error model is established by comparing the quasi-center of the distorted ellipse with the true projection of the object circle center. Then, an eccentricity error compensation framework is proposed which compensates the error by iteratively refining the image point to the true projection of the circle center. Both simulation and real experiment confirm the effectiveness of the proposed method in several vision applications.
1 March 2022 in IEEE Transactions on Industrial Informatics
Five-axis sweep scanning is an emerging technology, which can achieve precise and efficient contact measurement. However, existing five-axis scanning path planning algorithms either provide a feasible solution without considering measurement efficiency, or optimize the path without considering the interference-free condition. In this article, an algorithm is proposed for generating time-optimal and interference-free five-axis sweep scanning path. The measurement time is estimated by feed rate scheduling methods and used as the cost function of the optimization problem. The interference condition is determined by whether the stylus orientation is within the admissible orientation domain and formulated as the penalty function. The optimization problem is finally solved by the distributed particle swarm optimization to obtain the optimized scanning path. The experimental results validate the effectiveness of the proposed method. The proposed method has better measurement efficiency over the other existing five-axis sweep scanning path planning methods and a leading commercial software.
The normal vectors of the free-form surface profile curve are generally not in the same plane, even if the profile curve is generated from the interaction between a plane and a surface. If the 2D compensation vector is adopted to reduce the cosine error, the accuracy is limited because the normal direction of the curve plane is not considered. In this paper, a novel radius compensation method adopts the 3D compensation vector instead of the 2D compensation vector like most existing works. An approximating method is proposed to estimate the 3D compensation vector, which is composed of two parts. One is the projection of the 3D vector in the curve plane that is calculated by the normal vector of the non-uniform rational Bézier spline fit curve. The other is the normal vector of the curve plane passing the measuring point. In addition, the deviation between the workpiece and CAD model may degrade the accuracy of the radius compensation. Through theoretical analysis, the influence of the deviation cannot be ignored when the deviation is larger than 0.2 mm. By setting up this threshold, under various levels of torsion angles and deviations for the free-form surface, the results show that the compensation accuracy of the proposed method has a better performance than traditional 2D compensation methods, which decreases the maximum error from 70.8 to 2.6 μm.
11 September 2024 in IEEE Transactions on Instrumentation and Measurement
On-machine measurement (OMM) accuracy for freeform surface workpieces is greatly reduced by the error of the touch-trigger probe. Existing methods typically perform probe error compensation under the assumption of vertical measurement. However, due to possible workpiece deformation or unknown workpiece models, the approaching direction inevitably deviates from the normal vector of the inspection point, leading to non-vertical OMM. In this paper, the principal differences between vertical and non-vertical OMM processes are clarified first. Compared with vertical OMM, non-vertical OMM results in more complex probe pre-travel errors and introduces the problem of ruby ball radius compensation. Then, a novel touch-trigger probe error compensation strategy for non-vertical OMM is proposed. The pre-travel error of non-vertical OMM is divided into the delay error and the stylus deformation error. The former can be calculated by the approaching speed and OMM systems delay time. The latter could be figured out based on the 3D deformation error map and interpolation methods. The normal vector approximation of the inspection point is iteratively calculated by the B-spline surface fitting to compensate for the stylus deformation error and ruby ball radius simultaneously. For a designed freeform surface workpiece, the proposed strategy improves the non-vertical OMM accuracy by 49% compared with the comparative method.