According toResearch Question 1, the following key objectives that will help to satisfy the aforementioned research gap to select the robust laser dimpling process parameters to achieve given quality requirements in the presence of process variation.
• Objective 1: To understand the quality requirements of a dimple:
In essence, the dimple works as a spacer between two sheet metals to create a joining gap in which the zinc vapour is ventilated through. According to the revised literature given in Section 3.1, the only studied key performance indicator (KPI) is the dimple height which directly affects the joining gap size. However, there is a continuous clamping force acting upon the dimple during
welding. Thus, the dimple upper spot area can be another KPI to assess the strength of a dimple to prevent excessive deformation of the dimple height un- der compression of clamping force. Furthermore, the excessive amount of laser power creates a dark black spot on the lower surface of the sheet metal which degrades the surface finish. The dimple lower spot area can be considered as another KPI to assess the aesthetic quality of a dimple.
• Objective 2: To obtain process capability space considering a multi-input,
multi-output (MIMO) based scenario:
Laser welding is a complicated multi-phases and multi-physics process which involves interaction between the laser beam and material. This interaction is governed by a number of factors including laser power, laser intensity distri- bution; and process parameters such as scanning speed, incidence angle and focal offset. The proposed modelling approach addresses two key limitations as discussed in the literature by taking into consideration (i) approximation of comprehensive multi-variate relations between multi-input (i.e. process pa- rameters) and multi-output (i.e. key performance indicators), and(ii) process variation over the design space by introducing deterministic and stochastic pro- cess capability spaces. The deterministic process capability space is a measure of the dimpling process capability to satisfy simultaneously all the allowance limits of KPIs; whereas, the stochastic process capability space is the estima- tion of success rate (SR) which is the probability of making a dimple that satisfies simultaneously all the allowance limits.
• Objective 3: To find robust process parameters that are less sensitive to the
process variation:
The laser dimpling process is a pre-process for laser welding of galvanized sheet metals. It is important to note that the requirements of the laser dimpling process are determined by an upstream process, such as assembly fixture design and clamp layout optimization. For example, assembly fixture design for laser welding might require a specific value of KPIs with a given variability. In this case, deterministic and stochastic process capability spaces are utilized to find the robust process parameters that are less sensitive to process variation.
Based on the aforementioned objectives and the defined framework, the methodology for calculating surrogate driven deterministic and stochastic process capability space have been proposed inChapter 4. The research contributions of the proposed work are listed as:
• Contribution 1: The new KPIs and their intervals are defined to assess the dimple quality:
The product quality can be analysed by three features: (i) functionality,(ii)
durability, and (iii) aesthetic. According to the revised literature, one KPI, which is the dimple height, is studied to address only the functionality of the dimple. Two new KPIs (i.e. the dimple upper spot area and the dimple lower spot area) are introduced to address the remaining features. The dimple upper and lower spot areas are utilized the assess durability and aesthetic quality of the dimple quality, respectively.
• Contribution 2: A comprehensive characterization of the laser dimpling
process:
The industrial needs are addressed in this study by(i) introducing two new KPIs, and (ii) studying the effect of incidence angle and focal offset, which are required for accessibility issue of the laser beam as well as scanning speed and laser track, which are required for the cycle time. Furthermore, the de- terministic process capability space is introduced to find feasible the process parameters which simultaneously satisfy all quality requirements. Due to the stochastic nature of the process, the feasible process parameters might produce dimples that violate the allowance limits of KPIs. Therefore, the stochastic process capability space is introduced to measure the success rate using a prob- abilistic approach. Based on the desired success rate, the natural specification limits are determined to satisfy all quality allowance limits.
• Contribution 3: Process parameters selection and optimization using
surrogate-driven process capability space:
The proposed methodology offers a unique simulation tool which is generic and can be applied not only to the laser dimpling process but also it can be exploited in the context of selection and optimization of process parameters in the presence of process variation. The current best practice for process parameters selection is based on costly and time-consuming trial and error approach. The proposed methodology offers identification of risky areas and low reliable parameters settings which help to the selection of optimum process parameters and shorten the time for design and commissioning.
According to the Research Question 2, the following key objectives that will help to satisfy the aforementioned research gap to directly monitor multi Fusion Zone Dimensions (Penetration, Top Width, Interface Width) in overlap laser welding of galvanized steels to assess the weld quality
• Objective 1: To develop a fast model by decoupling occurring multi-physics in laser welding:
Laser welding is a complicated multi-phases and multi-physics process. As discussed, there is a requirement for monitoring multi FZDs to assess the weld quality. The lack of comprehensive models linking (i) in-process monitoring data (e.g. visual sensing, acoustic and optical emissions); with, (ii) multiple quality indicators (e.g. penetration depth, interface width), and (iii) weld- ing process parameters (e.g. laser power, welding speed, focal point position) underscores the limitations of current data-driven in-process monitoring meth- ods. Therefore, a model that gives results in a short time is required to predict the transverse cross-section of weld in which FZDs are obtained. To meet this objective, the sequential decoupled multi-physics model has been developed. The model calculates the keyhole profile in overlap joint using an analytic method; and then, solves the heat equation using FEM to obtain FZDs.
• Objective 2: To integrate experiment and numeric simulation results to im-
prove the accuracy of the model:
The term fidelity refers to the level of accuracy or complexity of the developed model. A simplified numeric simulation model (i.e. Low Fidelity (LF) model) of laser welding is developed by sequential solving the occurring physics. The key idea is to integrate the LF model with the experiment-based model (i.e. High Fidelity (HF) model) to reduce computational time without losing ac- curacy. For this purpose, two correction models are developed to increase the accuracy of the keyhole profile calculation in overlap joint. Then, heat equation is numerically solved based on the corrected keyhole profile.
Based on the aforementioned objectives and the defined framework, the methodology for calculating the decoupled multi-physics multi-fidelity (DMPMF) model is given proposed inChapter 5. The research contributions of the proposed work are listed as:
• Contribution 1: Development of sequential decoupled multi-physics model
considering incidence angle and joining gap:
Complex welding simulation models have a realistic estimation of the FZ pro- file as well as FZDs, but they are often computationally expensive. Inexpensive and less accurate the LF model can be achieved by dimension reduction, lin- earisation and considering simple physics. However, the LF model cannot be directly utilized for in-process monitoring because the output of the LF model (e.g. the FZ profile, FZDs, etc.) has a significant error. In the literature,
the influence of the incidence angle on the keyhole shape has not been fully addressed. In this work, the laser beam was assumed as a hyperboloid and the interaction of hyperboloid with any plane in the space was analytically cal- culated so that the laser beam on-surface shape and the laser intensity were obtained considering the incidence angle. According to the obtained intensity, the keyhole profile was calculated using a well-established analytical method.
• Contribution 2: Integration of correction model in to the LF model:
The multi-fidelity (MF) modelling method combines the information gain from both LF and HF models by using correction model. The correction model can be based on either (i) scaling factor (β) which is the ratio of the HF model results to the LF model results,(ii) discrepancy factor (δ) which is the differ- ence between HF and LF model results, or(iii) combination of both. After obtaining scaling and discrepancy factor, the correction model is generally de- veloped by employing surrogate modelling approach (i.e. regression, Kriging,
etc.). The MF model in this research is based on scaling factor to corrected calculated the keyhole profile based on the decoupled multi-physics modelling approach.