Elsevier

Computer-Aided Design

Volume 43, Issue 4, April 2011, Pages 345-355
Computer-Aided Design

3D part inspection path planning of a laser scanner with control on the uncertainty

https://doi.org/10.1016/j.cad.2010.12.014Get rights and content

Abstract

This article concerns the measurement process of mechanical parts using laser scanners. From the point of view of industrial applications, the objective is to guarantee the measurement accuracy during the scanning with regard to the geometrical product specifications. The proposed method can be summarized as follows: the first step consists of analyzing the interval of tolerance for the different specifications and to attribute to every geometrical entity a maximal uncertainty of measurement. This uncertainty depends on the angle of incidence between the laser plane and the scanned surface. In the second step, an approach based on the concept of visibility is used from the CAD model of the inspected part to find correct sensor guidance in a metrological point of view. A few position-points from this set are used to define the scanning path. Finally, the measurement can be carried out and the specifications can be controlled after the segmentation of the point clouds. An example illustrates the approach.

Research highlights

► We describe a method to find a good measurement trajectory of a laser sensor. ► From tolerance interval, the method permits to control the measurement uncertainties. ► The measurement ability of the sensor is a priori determined. ► The method can be automated and integrated in a C.M.M. inspection software.

Introduction

Today, the techniques of non-contact measurement, particularly the laser scanners, are used more and more in industrial devices like digitizing or reverse engineering of products. However, these techniques are less used in the inspection of mechanical parts with the objective of verifying dimensional or geometrical specifications. The reasons are: firstly, these techniques are less accurate than classical techniques with a contact probe for metrological applications and, secondly, there is no suitable methodology or aided tools to digitize complex mechanical parts by laser scanning. Indeed, digitizing paths are not given by registration of a few points attached to a surface but, due to the size of the field of measurement, several hundred points linked to one or more surfaces are measured simultaneously.

At the moment, there is no inspection aided tool to help measurement and verification of specifications by laser scanner, i.e. a tool enabling to determine a sensor path ensuring the accuracy of measurement.

Despite the accuracy of measurement of a few micrometers, verification by probing of dimensional and geometrical specifications also imposes some constraints. These constraints are: low speed scanning (the more complex is the measured part, the longer is the scanning), low acquisition rate of measurement points, reduced accessibility of certain surfaces like cylinders, difficult probing of flexible materials.

Non-contact measurement techniques, essentially optical techniques, allow a high resolution and rapid speed in the digitizing and measurement of non-rigid surfaces but they are less accurate than the measurement with contact probe. These techniques, most widely used in industrial applications are: photogrammetry, structured light and laser scanning. The photogrammetric technique is based on the observation of targets placed on the product with the use of several cameras. The triangulation principle using two image frames allows one to obtain the 3D coordinates [1]. The projection of structured light uses only one camera, the deformation of different patterns projected on the surface gives the three dimensions [2]. The laser scanning consists of observing the projection of a laser plane with a camera. In this case, it is necessary to move the sensor to observe the object. The specification measurement poses two kinds of problems to the industrial user. On the one hand, the measurement uncertainties are greater than those obtained with contact techniques. Prieto et al. have shown that, when using a laser scanner, the dispersion of measurement results of various geometrical and dimensional specifications can be four to ten times higher than those obtained by contact measurement [3]. Therefore, before accepting the product, it is necessary to check that the uncertainty is correct with regard to the interval tolerance of the specifications. On the other hand, it is necessary to associate every measurement point with their nominal geometrical entities and to check that all the surfaces have been measured.

Recent developments of 3D laser scanners, particularly in terms of accuracy, are of great interest in the case of the in-process product inspection with regard to industrial costs, quality improvement and complexity of manufactured parts. The approach of the product part inspection is different to the classical probe method with Coordinate Measurement Machine (CMM). The present work is based on the application of the laser scanner. However, the principles of the methodology can be applied to the other above mentioned techniques.

In this paper, the problem of verification of geometrical product specifications is considered. In particular, the scanning process is ensured with a laser plane device linked to a CMM (see Fig. 1). The originality of this work lies in the development of a strategy to determine a sensor path taking into account both the metrological properties of the sensor and the dimensional and geometrical specifications of the inspected part. These specifications decide on whether or not a real mechanical part is compliant. They are the main constraints for the search of a sensor path.

A lot of works involve the scanning of objects whose geometrical model is known or unknown. These works are generally intended to find a path that allows complete digitizing of the part with a few quality criteria of the point cloud. Among these criteria, indicators of density (ρ-dense) and completeness (κ-complete) defined by Hoppe [4] are used. Nevertheless, in a metrological context, these are insufficient and one has to take into account the uncertainties of measurement. These come from measurement operations and their estimation may be obtained following the methodology specified in the Guide to the expression of Uncertainty in Measurement (GUM) [5]. One can generally distinguish two types of uncertainty. The first one determines the systematic error which can theoretically be corrected. The other one determines the random error. Bourdet et al. [6] take into account these two parameters in the definition of additional criteria of quality of point clouds for a metrological control. These parameters are the accuracy (τ-accurate) and the noise (δ-noise). In the present work, a systematic errors model is used to correct the measurement error. So, the τ-accurate indicator is taken into account trough this correction.

Among the various works on scanning paths, different approaches can be described. Kweon and Meideros [7] proposed a methodology based on a visibility map, called VMAP (Visibility Map), for determining a set of positions and orientations to control specifications while limiting the number of manipulations subject to greater uncertainty of measurement on a CMM. However, the uncertainty is not taken into account. Son et al. [8] propose an automated measuring system for free form surfaces and whose CAD model is given. First, the part model is defined with a STL mesh while controlling the error of non-compliance using the curvature. Then, critical points are determined according to the curvature variations. A scanning path is generated by checking criteria of collision and size of the field of view. However, in this approach, the control of the uncertainty of measurement is not integrated. Bernard et al. [9] and Rémy et al. [10] were interested in the digitizing strategy linking direction and positioning of the sensor in relation to the object with the quality of the point cloud in terms of density and noise. From the concept of visibility, better sensor guidance is used. Here, the measurement noise is taken into account by limiting the visibility cone. Nevertheless, this limitation is done, a priori, without control of measurement uncertainty. More recently, Mehdi-Souzani et al. [11] proposed an intelligent scanning system based on quality control of the obtained point cloud. Its work concerns the scanning of an unknown object. Among these works, only Prieto [12] has proposed an acquisition strategy taking into account the specified tolerances with their numerical values. The proposed algorithm, built from the CAD model of the part, is as follows: create a voxel model of the part, determine position-points to measure the part, make sure there is no occlusion, estimate the accuracy of measurement data through a 3D model of measuring noise, and then calculate optimal position-points. The main drawback of this approach lies in the subsequent verification of the quality of measurement. On the contrary, our approach aims to find a path that guarantees, a priori, measurement uncertainties with regard to tolerances.

Recent works have been carried to control the measurement uncertainties with other optical sensors such as fringes projectors. Böttner have built a virtual fringe projection system (VFPS) to simulate the device defects, each model parameter being controlled individually and independently [13]. Weckenmann et al. [14] have also studied the best placement of the projector and the camera to obtain a minimum measurement uncertainty. A model based on the reflection light influence and surface roughness is carried out. This model maximizes the accessible surface zone and to minimizes the measurement uncertainty, this one depending on the incident and reflection angle of the light ray onto the surface. A complete algorithm is developed to support the inspection planner and to define the inspection task.

In the present work, the approach is similar. However, on one hand the goal is not to find a positioning but to generate a scanning path, on the other hand, it is focused on the overall approach of mechanical part inspection, notably on the specifications analysis. The acceptable device uncertainties are considered with the objective to guarantee a measurement with a suitable ratio of uncertainty to tolerance. Thus, taking into account some influent parameters and indicators describing quality of the point cloud, it is possible to generate a scanning trajectory according to the considered specifications. The measurement problem can be summarized as following:

  • 1.

    preliminary analysis of the part and its specifications,

  • 2.

    estimation of an acceptable uncertainty of measurement in comparison with specifications,

  • 3.

    the search for the set of possible orientations and positions of the scanner in respect of the expected limit uncertainties,

  • 4.

    generation of a scanning path,

  • 5.

    control by comparison of part deviations with specified limits.

This must be done according to Geometrical Product Specifications (GPS) standards [15], [16] and measurement procedures.

Section snippets

Acceptance criteria of mechanical part

As part of ISO (International Organization for Standardization), different works have been carried out on the geometry of part and the associated requirements. The objective of these works is, through a set of standards, to provide dimensioning and geometrical tolerancing as a symbolic language used to explicitly describe nominal geometry and its permissible variation. This language, called GPS, can be read and understood by all the actors (designer, manufacturer and controller) operating in

Example

The considered part is a part that insures a positioning functionality. Eight geometrical entities (seven planes and one cylinder) characterized by different dimensional specifications are considered: distances between two parallel planes called L1 and L2, diameter D1 associated with hole C1. Two other specifications are relative to geometrical specifications: tolerance of perpendicularity of hole axisAC1 related to plane P7 and angular tolerance of plane P5 to plane P6. All of these

Conclusion

In this paper, a metrological approach for dimensional and geometrical inspection of mechanical products with a plane laser sensor and a CMM has been presented. The originality of this work is based on initial analysis of part specifications before inspection taking into account uncertainties of measurement of the laser scanner. For each step of the process, the quality of the measurement is considered through three criteria: density, completeness and accuracy of measurement. Then, the problem

References (26)

  • ISO/IEC Guide 98-3:2008 — part 3: guide to the expression of uncertainty in measurement (GUM: 1995);...
  • P. Bourdet et al.

    Quality of 3D digitized points obtained with non-contact optical sensors

    Annals of the CIRP

    (2002)
  • A. Bernard et al.

    Définition des paramètres essentiels pour la génération automatique de processus d’acquisition de forme par capteur laser

    Revue Internationale de CFAO et d’Informatique Graphique

    (1998)
  • Cited by (0)

    View full text