Efectos de la Reforma constitucional de 2013

8.2.3 Summary of quality assessment framework

This section has discussed the integration and application of the newly proposed shape characterisation tools. This provides a dimensional quality assessment framework that can potentially allow manufacturers to more accurately define and assess assembly quality, which can lead to the ultimate goal of optimal process design for dimensional control. The next sections will now present the key contributions of this thesis along with areas for future research.

8.3

Contributions

Sheet metal assembly is a complex process involving component-to-component and component-to-tooling interactions. A key characteristic of sheet metal assemblies, the exibility of components, means that variation does not stack-up according to the addi- tive theorem of variance that applies to rigid bodies. Instead, components can be bent and distorted into conforming or non-conforming shapes by assembly interactions. This characteristic of exibility also means that in comparison to rigid body assembly, ad- ditional aspects of the assembly process, such as clamp sequence and weld sequence, can in uence the way in which variation propagates. In order to accurately investigate and model variation stack-up in sheet metal assemblies, the mechanistic behaviour of the assemblies must be taken into account. In this thesis, variation propagation in sheet metal assemblies is firstly investigated through a combination of finite element simulation and experimentation. It was shown how different clamping sequences can change the dimensional outcomes in terms of mean shape and variability, and therefore dimensional quality. It was also shown how clamping sequences could be adjusted to target particular quality assessment criteria. However, it has been discussed how exist- ing univariate measures of assembly quality such as mean, standard deviation, process capability, and SPC methods, are perhaps not the best measure of assembly quality. This is primarily due to their inability to adequately capture a key characteristic of assemblies: correlated variation patterns. This thesis proposes that assembly quality cannot be simply assessed by the mean and variance of a set of assumedly statistically independent measurement points, and that correlated variation patterns in the form of bows, buckles, twists and ripples also form a large part of assembly quality perceptions. In order to truly optimise a process for dimensional quality, correlated variation pat- terns must therefore be taken into account. This leads to the research question posed at the beginnig of this thesis:

How can correlated variation patterns in sheet metal assemblies be char- acterised in ways that can lead to more advanced levels of dimensional qual- ity assessment and control?

136 CHAPTER 8. CONCLUSION

In addressing this question, two key shape characterization methods were proposed: the multivariate statistical shape model, and the local shape descriptors. These shape charaterisation measures were designed to overcome key limitations of existing univari- ate quality measures including an inability to capture correlated variation patterns, monitor non-normally distributed data, interpret high dimensional data, and measure local variation patterns of different sizes or scales. Through addressing these limitations a more accurate description of assembly variation is achieved, allowing for improved quality assessment practices. Investigation of the sheet metal assembly process and development of the new shape characterization measures resulted in several main con- tributions: the clamp sequence simulation approach, clamp sequence design, the mul- tivariate statistical shape model, the local shape descriptors, and a quality assessment framework.

8.3.1 Clamp sequence simulation

In this thesis a non-linear finite element contact model implemented in a commer- cial code was adopted to provide a detailed representation of the physical sheet metal assembly clamping process. An advantage of this approach over commonly used lin- earized approaches is that it can capture contact resulting from the initial placement of components in assembly fixtures, and for deformation induced contact. Linearized approaches require contact points to be pre-specified, which often means that only weld points and clamp locations are assigned to come into contact: other surface locations can therefore effectively pass through each other as they have not been specified as contact locations. Efforts have been made to estimate possible deformation induced contact points, however, non-linear contact models still provides a more detailed repre- sentation of the process. A key disadvantage of nonlinear contact models is that they are computationally intensive, and therefore are not easily combined with Monte Carlo simulation for variation propagation analysis.

8.3.2 Clamp sequence design

This thesis highlights the possible extent of the in uence of clamping sequence on both mean shape and variability, and how clamp sequence can be altered for the dimensional control of different variation patterns. Furthermore, a set of generalised clamp sequence design laws is developed to provide a practical guideline for future process design, which could be readily integrated within increasingly popular knowledge based design systems.

8.3.3 Multivariate statistical shape model

The multivariate statistical shape model (KDE-PDM) was developed to provide a more advanced statistical representation of assembly process measurement data, enabling

8.4. SUGGESTIONS FOR FURTHER WORK 137