Search

A review of several statistical methods that are currently in use for outlier identification is presented, and their performances are compared theoretically for typical statistical distributions of experimental data, considering values derived from the distribution of extreme order statistics as reference terms. A simple modification of a popular, broadly used method based upon box-plot is introduced, in order to overcome a major limitation concerning sample size. Examples are presented concerning exploitation of methods considered on two data sets: a historical one concerning evaluation of an astronomical constant performed by a number of leading observatories and a substantial database pertaining to an ongoing investigation on absolute measurement of gravity acceleration, exhibiting peculiar aspects concerning outliers. Some problems related to outlier treatment are examined, and the requirement of both statistical analysis and expert opinion for proper outlier management is underlined.

Current standard procedures for the evaluation of measurement uncertainty of Coordinate Measuring Machines (CMM) being not fully satisfactory, other methods are considered. Conformity to geometric and dimensional tolerances, specified for an industrial workpiece previously involved in a comprehensive round robin test, was checked with CMM. Measurement uncertainty was also assessed exploiting a simulation method, developed at Politecnico di Torino, offering several advantages in terms of CMM operability and substantial cost savings. Estimates of measurement variability due to single and combined effects of factors considered, obtained from testing in a representative verification, are discussed in the light of results of simulation.

Quality may be defined as a set of requirements a system should satisfy in order to meet customer’s needs. Control of these requirements assures satisfaction of relevant standards, and consequently the performance levels of a manufacturing/transactional stream. In this context it is fundamental to define control procedures and reliable measurement systems adequate for adopting improvement action as soon as anomalies and dysfunctions are detected. This paper deals with a study of measurement variability occurring during practical exploitation of CMMs (Coordinate Measuring Machines). These measurement systems are designed to probe selected points of workpiece surface, and compare the relevant coordinates or derived quantities with specified values; capability and versatility of CMMs justify their widespread use in industry. Evaluation of CMM measurement variability is however often awkward owing to a number of factors, such as e.g. measurement task, environment, operator and measurement procedures. A round robin exercise involving two industrial laboratories was planned in order to address these issues. Three typical machine tool parts were circulated among participants, who were asked to measure linear dimensions as well as tolerances at specified locations, according to an agreed upon schedule. Results of measurements, performed by experienced CMM industrial users, were analyzed in order to bring out discrepancies, and suggest remedial actions in the light of information gathered. Several factors involving metrological as well as other aspects were observed to cause major discrepancies, yielding in turn information on where to look for potential sources of trouble. Conclusions were drawn in terms of operating procedure, leading to improved information on origin and components of variability.

Classical sensitivity testing addresses mainly problems where the level of one stimulus only governs an abrupt transition in output, or response. Both parametric and nonparametric methods developed, and successfully applied over last century to tackle such problems, provide estimates of critical levels beyond which an item will either respond, or not, to a single stimulus, and of related statistics. However classical methods sometimes may not readily provide an answer, namely when more than one stimulus may reach critical level, and either singularly or jointly trigger transition. Factorial and response surface designs, adequate when dealing with continuous responses, may not perform as well for transition threshold estimation. A practical case at hand in chemical engineering concerns the production, through hydrolysis of a specific precursor, of titania sols and gels that find industrial use as additive for paints, concrete and other building materials due to its optical, photo-catalytic and super-hydrophilic properties. Particles formation and aggregation — controlled by varying the primary process parameters, namely initial alkoxide concentration, water to alkoxide and acid to alkoxide ratios, mixing conditions — may yield either stable, transparent nanometric sols, or monolithic gels, where aggregation of nanometric particles produces a final ceramic object. Depending on the application, one of the two products may be desirable, and therefore it is crucial to control the final product properties. Aggregation kinetics and physical properties of sols, and sol to gel transition, were found to depend strongly upon several factors, that is water to alkoxide initial concentration ratio, acid to alkoxide initial concentration ratio, and their interaction. The approach developed in order to estimate parameters pertaining to transition, and related uncertainty, is presented in the paper, and discussed in the light of experimental results.