0
Research Papers

Using Predictive Modeling and Classification Methods for Single and Overlapping Bead Laser Cladding to Understand Bead Geometry to Process Parameter Relationships

[+] Author and Article Information
R. J. Urbanic

Department of Mechanical, Automotive,
and Materials Engineering,
University of Windsor,
401 Sunset Avenue,
Windsor, ON N9B 3P4, Canada
e-mail: jurbanic@uwindsor.ca

S. M. Saqib

Department of Industrial and
Manufacturing Systems Engineering,
University of Windsor,
401 Sunset Avenue,
Windsor, ON N9B 3P4, Canada
e-mail: saqibs@uwindsor.ca

K. Aggarwal

FCA,
800 Chrysler Drive,
Auburn Hills, MI 48326
e-mail: kush.aggarwal@external.fcagroup.com

1Corresponding author.

Manuscript received December 30, 2014; final manuscript received November 27, 2015; published online January 4, 2016. Assoc. Editor: Z. J. Pei.

J. Manuf. Sci. Eng 138(5), 051012 (Jan 04, 2016) (13 pages) Paper No: MANU-14-1721; doi: 10.1115/1.4032117 History: Received December 30, 2014; Revised November 27, 2015

Developing a bead shape to process parameter model is challenging due to the multiparameter, nonlinear, and dynamic nature of the laser cladding (LC) environment. This introduces unique predictive modeling challenges for both single bead and overlapping bead configurations. It is essential to develop predictive models for both as the boundary conditions for overlapping beads are different from a single bead configuration. A single bead model provides insight with respect to the process characteristics. An overlapping model is relevant for process planning and travel path generation for surface cladding operations. Complementing the modeling challenges is the development of a framework and methodologies to minimize experimental data collection while maximizing the goodness of fit for the predictive models for additional experimentation and modeling. To facilitate this, it is important to understand the key process parameters, the predictive model methodologies, and data structures. Two modeling methods are employed to develop predictive models: analysis of variance (ANOVA), and a generalized reduced gradient (GRG) approach. To assist with process parameter solutions and to provide an initial value for nonlinear model seeding, data clustering is performed to identify characteristic bead shape families. This research illustrates good predictive models can be generated using multiple approaches.

Copyright © 2016 by ASME
Your Session has timed out. Please sign back in to continue.

References

Figures

Grahic Jump Location
Fig. 1

Process planning process flow for laser cladding

Grahic Jump Location
Fig. 2

LC apparatus—schematic view

Grahic Jump Location
Fig. 3

Clad beads (a) single bead, (b) three bead overlay, and (c) overlap geometry [Courtesy of industrial Partner]

Grahic Jump Location
Fig. 4

Clad bead nomenclature for the (a) single and (b) overlap bead configurations [17]

Grahic Jump Location
Fig. 5

Process flow map for the methodology

Grahic Jump Location
Fig. 6

Experimental methodology overview

Grahic Jump Location
Fig. 7

K-mean clustering approach—schematic diagram for the algorithm

Grahic Jump Location
Fig. 9

Summary of significant factors for the predictive models, showing the number of square and interaction factors on the predictive models

Grahic Jump Location
Fig. 17

(a) The observed area versus focal length for the single bead experiments and (b) impact of the contact tip to work piece distance on height with constant power, powder feed rate, and travel speed conditions

Grahic Jump Location
Fig. 15

The R2 values for the bead height for the overlapping bead experiment set

Grahic Jump Location
Fig. 16

The R2 values for the bead penetration for the overlapping bead experiment set

Grahic Jump Location
Fig. 14

The R2 values for the bead width for the overlapping bead experiment set

Grahic Jump Location
Fig. 10

(a) Predicted overlapping bead width geometry versus observed bead width geometry and (b) standard deviations

Grahic Jump Location
Fig. 11

The average penetration and height changes with the percentage overlap and representative bead with 60% overlap illustrating the height and penetration variations per bead

Grahic Jump Location
Fig. 12

The average penetration and height changes with the percentage overlap

Grahic Jump Location
Fig. 13

The penetration variations based on the bead deposition order and percentage overlap

Grahic Jump Location
Fig. 18

Width, height, and penetration centroid distances for (k = 5) clusters, with the grand centroid for each shape element represented by phantom lines

Grahic Jump Location
Fig. 19

Silhouette plot for (k = 5) clusters

Grahic Jump Location
Fig. 23

Representative bead shapes (a)–(j) for k = 10 optimal clusters

Grahic Jump Location
Fig. 8

Bead width model results versus observed data (a) when applying a Log10 data transformation on the input parameters (b) using the ANOVA model

Grahic Jump Location
Fig. 20

Dendrogram plot for (k = 5) clusters illustrating the average linkage/Euclidean distance for the collected bead geometry data

Grahic Jump Location
Fig. 21

Dendrogram plot for (k = 5) clusters illustrating the average linkage/Euclidean distance for the collected bead geometry data and representative beads illustrating these cluster conditions (a)–(e). Note the differences in cluster (c) and (d). The penetration is deep, but the deposited material is minimal for cluster (c), whereas the opposite is evident for cluster (d).

Grahic Jump Location
Fig. 22

Calinski–Harbasz Scree plot for the LC single bead experiment

Tables

Errata

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In