universal method to optimise measurement uncertainty, time and cost for CMM scanning technology Prof. Dr.-Ing Robert Schmitt, Dipl.-Ing. Susanne Nisch Laboratory for Machine Tools and Production Engineering WZL RWTH achen University Steinbachstraße 19, 52072 achen, Germany Tel.: +49 (241) 80-27112 Fax: +49 (241) 80-22193 S.Nisch@wzl.rwth-aachen.de bstract Today scanning technology on coordinate measuring machines (CMM) is normally used for tactile detection of form deviations. Many companies apply CMMs instead of form measuring devices because of their high flexibility and the possibility of faster workpiece inspection. However, the kinematic of CMM is more complex compared to specialized form measurement devices. Dynamic effects like centrifugal force in roundness measurements results in a rising measurement uncertainty. The optimum setup of scanning parameters like point density and scanning speed is not easy and depends on the machine operator. Mostly, the dynamic characteristics of the used CMM are not sufficiently identified and the detection of the measuring task dependent uncertainty is very complex. t the Chair for Production Metrology and Quality Management of RWTH achen University an approach for optimisation of scanning parameters is developed. Therefore a CMM-specific scanning performance base is used, containing basic information about the used CMM, ranges for parameters and the main dynamic characteristics based on experiments. Keywords: Scanning on CMM, measurement uncertainty, efficient measurement 1. Instruction The application of industrial production metrology is always connected to the right consideration between the most efficient way of use and the targeted measurement uncertainty. The choice of the best measuring strategy is complicated for processes with nonlinear correlations between parameters and uncertainty because of rising measurement durations and costs. This paradigm can also be applied to scanning technology on tactile coordinate measuring machines [1]. There are multitudes of parameters like scanning speed, point density and force that influence the dynamic characteristics of coordinate measuring machines and the friction between the workpiece surface and the probe [2]. During data processing, outlier elimination and filtering, the resulting measuring deviation change in a non-linear way and the coherences between parameters, specific machine characteristics and measuring results cannot be quantified. ccurate specifications for parameters that allow a qualified and human independent choice of adjustment do not exist because of their dependence from workpiece surface and specific characteristics of every measurement device [3]. The right adjustment of parameters is complicated and the results of different CMMs are not comparable because of different conditions during the measuring process. lso the evaluation of measurement uncertainty for form measurement on CMMs does not exist. The characteristic limit value for form deviation MPE RONt only considers statistical variances in roundness measurement [1]. However the description of quantified coherences in the scanning process and data processing is not possible. Therefore the development of a method to choose the value of parameters by considering all significant conditions and maximum tolerable measurement uncertainty is necessary. 1-051
2. nalysis of parameters ccording to ISO 1101 the form deviation is evaluated by two concentric circles with a minimal radial distance. Measurement results are only partially influenced by measurement errors because the evaluation of form deviation only uses extreme values. Hence, this method is not capable for analysis of coherences between parameters and results. For detection of the real influences of parameters the Fourier transformation and calculation of dominating structures on the surface is well suitable. Fig. 1. Schematic version of the wave standards with the diameter of 80 mm and structured areas for inner roundness (1), outer roundness (2) and flatness (5) and areas for justification on form measurement devices (3, 4) and CMM. Using multiwave standards with different superimposed sinusoidal waves shown in Fig. 1 the real profile can be detected by calibrating on a form measuring device with high accuracy. If the Fourier transformation is used for analysis of calibration data of multiwave standards the profile of dominating waves DFT {FDom} is detectable. nalyzing a profile measured on a CMM the Fourier transformed profile looks different to the calibrated because of a super positioned error profile called noise noise. For the analysis of effects at measurement process, forcing errors at the measured profile total have to be separated by using equation (1). noise total Dom 1 DFT F (1) q, noise 1 n n i 1 2 noise, i (2) The noise signal enables the evaluation of the reliability when transferring particular parts of the profile depending on values of the input parameter. Using the statistical characteristic square middle of the noise (2) the measuring results are comparable. In addition the square middle is used for evaluation of measurement uncertainty. Performing a test series with n measurements under the same conditions the variance of square middle of noise enables the evaluation of random influences against the used values for scanning parameters. further application is the evaluation of maximum endurable values for scanning parameters like scanning speed. On condition that useful results for form deviation are obtained, the limits for scanning parameters have to be detected. By the application of the method Design of Experiments (DoE) the basic characteristics of scanning on coordinate measuring machines are analyzed and quantified systematically. With preliminary experiments the significant parameters were identified. Necessary information about limits of the parameters is used for definition of experimental limits. For example, the maximal possible scanning speed of a CMM that can measure 450 points per second is 45 mm/s with a point distance about 0.1 mm. Therefore the maximum possible scanning speed is limited depending on the required point density and the detected limit by keeping useful results. 1-052
Because of the high number of important parameters, full factorial experimental designs are not useful. They would require much time for the experiments and analysis without benefit compared to reduced experimental designs. Not all results are significant and important for analysis of errors of scanning processes. There are several interactions between some parameters, but the main interesting concern the scanning speed. The main goal while preparing the experimental design is to reduce experiments by mixing experiments of interacting parameters with experiments of non-interacting parameters. On the basis of preliminary experiments significant interactions are detected that are comparable on different machines. They only vary in value and direction. Based on the researches reduced experimental designs are available that only have to be adapted to a relevant target. 3. Coherences between measurement uncertainty and scanning speed The time for form measurement is mainly influenced by the scanning speed. Depending on other parameters like point density, measurement force and probe configuration the maximum useful scanning speed can vary. Hence, the evaluation of coherences and interactions with the scanning is essential. The optimal adjustment of parameters with scanning speed is not easy because of their dependence from surface and dimension of the workpiece. E.g. measuring a grinded workpiece with small diameter, less points are necessary and a faster scanning is possible than for measurement of a big workpiece with a rudely cut surface. Fig. 2. Measuring of form deviation (roundness) according to ISO 1101 with different scanning speeds and illustrating Fourier transformed profile for speed of 1 and 40 mm/s. In Fig. 2 an example is given for explanation of coherences between scanning speed and form deviation. Because of dynamic machine behaviour the form deviation increases with faster scanning according to the calibrated form deviation of the multi wave standard. For two speeds, slow scanning with 1 mm/s and fast scanning with 40 mm/s, the measured profiles are evaluated exemplarily. In the pictured profile it is well visible that the faster scanning causes a higher noise resulting in a higher form deviation. Evaluating the dominating waves via Fourier transformation the measured error is clear visible. With rising scanning speed the amplitudes of the 5 dominants are reduced and the amplitudes of other frequencies rise. This effect intensifies with rising frequency and depends on the point density of the measured profile. When the scanning speed reaches the critical value the probe is partially lifted from the workpiece surface because of the centrifugal force, surpassing the measuring force interacting in the reverse direction. This results in a smaller form deviation despite a bigger measuring error. 1-053
4. Establishing a model for form measurement For a systematic evaluation of best fitting parameters, the implementation of influences and correlations in a practical model is required. This model has to include the process of detection as well as the data processing and filtering because of the significant transformation of data. For definition of transfer characteristics the flow chart of scanning process on the left side of Fig. 3 is analysed concerning the transfer elements. The coherences between input and output are not detectable for every single transfer element. Hence, the process is abstracted to one total transfer element called transfer function in the model, shown in the middle of Fig. 3. n ideal input profile is filtered by morphological procedures to simulate the touch of the probe tip on workpiece surface. fterwards the process of touch detection and signal acquisition is simulated by the transfer function. This transformation bases on a super positioning of the filtered profile with an error profile. Random influences are considered by super positioning the transformed profile with a white noise. Both the error profile and the parameters of white noise are evaluated from the experimental detected noise profiles. Fig. 3. Flow chart for real scanning process, layout of the model for analysis and simulation of the form measurement on CMM and the configuration of the scanning performance base. In form measurement the row profile is processed to minimize the influences of outliers and roughness on the results. In the model this is processed in part two. The influence of data processing like outlier elimination and filtering on the measurement uncertainty is considered. By constant scanning speed the single influences could vary in specific limits and with differing probabilities. Therefore the possible combinations should be simulated. From the variance of the possible values the real measurement uncertainty of scanning process is evaluated. 5. pplication of a CMM-specific scanning performance base The basis of an object-oriented and uncertainty dependent adjustment of scanning parameters is the quantification of the coherences between influencing parameters and measured profile. The challenge is the evaluation of the transfer functions for simulation of scanning process. Coherences and interactions are identified and integrated into the scanning performance base, shown on the right side of Fig. 3. Beside the detected limits for variable parameters, CMM-specific non-adjustable parameters and necessary information about e.g. most stable probe configurations, the transfer functions are obtained. 1-054
The transfer characteristics depend on the individual CMM. However, the dynamic characteristics of CMMs of same type are very different. Hence, the specific detection of the transfer characteristics of every CMM for the tactile scanning is required. 6. Conclusion In the paper methods are presented for the experimental based optimization of scanning processes. The transfer characteristics of scanning processes support the evaluation of measurement tasks specific uncertainty for form measurement. The main goal is the evaluation of scanning parameters to keep the tolerable scanning speed. The evaluation of scanning speed dependent measurement uncertainty bases on simulation. Further work has to focus on the development of algorithm for adjustment of scanning speed to the maximum tolerable measurement uncertainty depending in specification for dorm deviation. Previous experiments showed a strong correlation between the detected influences at scanning as well as the dimension and surface characteristics of the workpiece. The achieved results and coherences are not transferable to bigger workpieces or other structured surfaces. t present, there are many restrictions to the use of this method. dditional analysis will serve to expand the method to any workpiece dimension and surfaces. Therefore experiments with real workpieces from different manufacturing methods will be analyzed. 7. cknowledgements The authors would like to thank the Deutsche Forschungsgemeinschaft DFG (German Research Foundation) for the support of the depicted research. References 1.. Weckenmann, B. Gawande. Coordinate Measurement Methods Flexible Measurement Strategies for Measurement, Form and Position, Koordinatenmesstechnik Flexible Messstrategien für Maß, Form und Lage. Hanser Fachbuch, München. 2007. 2. O. Jusko, F. Lüdicke and F. Wäldele. High Precision Form Measurements with Coordinate Measurement Machines. Tagungsbund zum X. Internationalen Oberflächenkolloquium, Chemnitz. 2000, pp. 341-351. 3. T. Pfeifer,. Napierala. Scanning on coordinate measuring machines. XVI Imeko World Congress-IMEKO 2000, Vienna, ustria. 25-28 September 2000. 4. J. Seewig, T. Hercke, N. Rau, M. Mills, M. Meyer, R. Volk and H.-J. Kedziora. Dominant Waviness a practice oriented procedure for waviness evaluation. Tagungsband zum XI. Internationalen Oberflächenkolloquium, Chemnitz. 2004, pp. 198-207. 5. W. Lotze. High Speed Scanning on Coordinate Measurement Equipment. High-Speed Scanning auf Koordinatenmessgeräten, Microtecnic. 1993, 4. 6. E.O. Brigham. FFT Fast Fourier Transformation, FFT Schnelle Fourier- Transformation. 6th Edition, Oldenbourg, München. 1995. 1-055