Comparing optical coordinate measuring machines reveals the need for standardization

Optical dimensional measuring techniques are increasingly used for quality control across many industries due to their high speed and other inherent advantages of contactless techniques. However, compared to tactile measuring devices, these sensors have additional error sources and a lack of standard verification procedures. Currently, there are no international standards for testing the metrological performance of optical coordinate measuring machines (CMMs). Only a few national standards or guidelines are available, such as those set by the German Institute for Standardization (e.g., DIN 32877), and the Association of German Engineers and German Association for Electrical, Electronic, and Information Technologies (e.g., VDI/VDE 2617-6 and VDI/VDE 2634). Due to these limitations, industrial users often find it difficult to evaluate the actual metrological performance of these systems.

In this context, we decided to conduct an industrial comparison of 21 optical CMMs from various companies in Europe (Italy, Switzerland, Spain, and Denmark). The study, called VideoAUDIT, was held between August 2007 and January 2009.¹ The participants measured several items, which we periodically calibrated in our pilot laboratory. For each item, we designed a detailed measuring procedure, which garnered a great amount of data from the measurements. The principal aim of the comparison was to provide the participants with calibrated objects and procedures for verifying the actual metrological performance of their optical CMMs. Secondly, we wanted to evaluate the quality of optical measurements and the ability to calculate measurement uncertainties in industrial product inspection.

Figure 1. Audit items: glass scale (1D), hole-plate provided by the Technical University of Denmark (2D), and four polymeric bricks (3D).

Participants measured a set of calibrated items at different levels of complexity. These items represent standards for metrological performance verification and are common industrial products (see Figure 1). The stability (e.g., no change in dimensions and geometry over time) of the items was checked several times during the circulation period, repeating the calibration procedures at regular periods. The stability tests showed good results. In particular, the calibration procedure for the polymeric bricks (which are the least stable items) showed that the maximum measurement variations were within 2μm for lengths and 3μm for geometrical features, which is considered acceptable for the purpose of an industrial comparison.

We analyzed all of the measurement results. For each measured dimension, we computed the deviation from the reference value and the normalized error (the so-called E_N-value, which combines the deviation and uncertainty of the CMM measurement into one number. An E_N<1 indicates good agreement between the two results.) Compared with the 2D and 3D items, the glass scale was easily measured by the participants. The task in this case was to measure 10 bidirectional lengths, following the procedure indicated by draft International Organization for Standardization (ISO) 10360-7.² Figure 2 shows the measurement deviations from the reference values and related uncertainties of the CMM measurements for three selected nominal lengths: 30mm (L30), 120mm (L120), and 210mm (L210). Analyzing the results provides important feedback about the presence of systematic errors (such as geometrical errors of the machines axis or ones due to ineffective correction of temperature effects) for each CMM. Measurement results from the 2D item, a hole plate provided by the Technical University of Denmark,³ show larger deviations than those from the 1D item, primarily due to the increased complexity of the measuring task. Table 1 shows the E_N values obtained by the participants who measured the hole plate. A better understanding of the errors can be obtained by reviewing the deviation of each single hole presented in Figure 3, which shows the squareness error for the machine axes of one of the participants. An interesting result is that participants experienced more difficulty measuring form errors (such as roundness) than they did with dimensional measurements.

Figure 2. Glass-scale (1D) measurement results.

Figure 3. Example of squareness error based on the measurement of the 2D item by one of the participants.

In the case of the polymeric 3D items, higher complexity measuring tasks led to a larger dispersion of results, particularly when compared with the similar length L30 on the glass scale (see Figure 4). For example, only six participants obtained E_N<1 for length measurements on the white brick.

Figure 4. Measurement of sizes on plastic bricks. Error bars refer to the uncertainties stated by the participants.

In conclusion, the VideoAUDIT comparison^4,5 shows that industrial optical CMMs need better periodic verification with suitable standards and procedures. The comparison helped to improve the measuring capability of industrial participants, revealing a variety of error sources that can be eliminated or taken into account in determining uncertainties. Further results of the comparison will be presented in future publications.