How do we ensure the robustness of the links between specification and evaluation in data measurement analysis and treatment ?

Table of contents Contract ................................................................................................................................................... 2 Robustness .......................................................................................................................................... 3 Correlation ........................................................................................................................................... 3 Measurement & evaluation ................................................................................................................ 3 Situation / assessment ............................................................................................................................ 3 Multicriteria & multidimensional matrix ............................................................................................. 3 Detail of each criteria .......................................................................................................................... 4 Needs & specification ...................................................................................................................... 4 Device / Means ................................................................................................................................ 4 Software .......................................................................................................................................... 4 Method ............................................................................................................................................ 4 Treatment & analysis ....................................................................................................................... 5 Distortions ........................................................................................................................................... 5


Robustness
The robustness of measurement & evaluation must be demonstrable. It takes into account the uncertainty of measuring devices and the bias generated by the translation of needs, measurements, the data treatment, the final analysis and transcription. This is a major issue for dimensional and geometrical measurement. Results have to be published and we must be able to measure the overall uncertainty of the activity.

Correlation
The control report must include all the elements necessary for the operations of the report's applicant (manufacturing or R&D), these elements must be clear, comprehensible and exploitable by nonmetrologists.

Treatment & analysis
Each criterion, each adjustment, each choice increases the multiplicity of the cases. This multiplicity is the major factor of the measurement diversity (i.e inter-laboratory comparison)

Some examples • Need translation onto specifications
The ISO technical language (ISO 8015, 1101…) is complex and closed. It requires a deep analysis and a detailed transcription of the datum reference systems. This language is specific to the functional compliance analysis of the parts (components). It makes the transcription of other needs difficult; i.e report for tool adjustment.
Example: a wrong transcription of the datum reference systems distorts all the position evaluations • Software Software packages should « filter » the measured cloud of points. These filters require adjustment, of which the metrology lab is in charge. It is difficult to assess the impact of these adjustments on the accuracy of the evaluations relative to the measured part.
Example: for laser scanning, do we transform the cloud of points into a polygonal model? The "noise" will be very different • Devices / means A same surface measured by different types of devices (different technologies) may have divergent evaluated characteristics.
Example: Measurement with laser scanning then contact probing: the noise will be significantly different. It generates disparity • Inter-laboratory comparison A significant example of distortion is the inter-laboratory comparison. Considering device measurement uncertainty and a context of simple specifications, it is difficult to obtain similar (close) values for one part. The difference is often larger than the uncertainty of measurement.
Example: measurements done inside the company and the one's done in a supplier lab.

Interdependence of criteria
Finding interdependence for criteria is a difficult path. Only a few common cases can be treated The multiplicity of criteria and their biases makes it impossible to write standard rules that cover them in their entirety.

links specification -method
Because of the criteria multiplicity and their values, it is rare to be able to create logical links between specifications and measurement method (including value adjustments)

Existing solutions
The existing solutions, demonstrated as robust, feed the return of experience.

Solution GEODESIGN expertise
Collecting, analysing, managing the criteria multiplicity

Scripts / drawings
The transcriptions are analysed to guarantee the link between need and specification

Return of experience
Each "solution" is analysed in order to extract a model or part of a model Capitalisation can be in the form of libraries.

Facilitator
An analysis hub is created. Its goal is to collect, analyse and manage the different parameters & criteria then to give recommendations for measurement & evaluation