Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
Master Thesis
Accuracy Control and Welding Distortion Prediction in a Deck Plate
Marcio Fleming7th EMship cycle: October 2016 − February 2018
Supervisor: Prof. Remigiusz Iwańkowicz, West Pomeranian University of Technology, Szczecin, PolandCo-supervisor: Prof. Jean-David Caprace, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil
Internship tutor: M.Sc. Nicole Schenk, Fr. Lürssen Werft GmbH & Co. KG, Bremen, GermanyReviewer: Prof. Hervé Le Sourne, Institut Catholique d’Arts et Métiers, Nantes, France
La Spezia, February 2018
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
2 of 9
1. Motivation
2. Methodology
3. Model
4. Results and Analysis
5. Conclusions
6. Future Works
Contents:
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
• Welding Process Parameters:• heat input;• welding sequence;• preheating/post heating;• gaps;• tack welds;• the edge preparation;• welding conditions;• interpass temperature;• the shape of penetration;• positioning;• welding procedure;• and the degree of restraint during welding.
Buckling Bowing Angular
Out-of-Plane
Transversal Longitudinal
3 of 9
1. Motivation
Wide Studied - Different Types of Distortions - > 1940
Influence Parameters
Non linear process
Machine learning exploration opportunity
In-Plane
• Geometric Parameters• Dimensions of the structure• Type and size of welded joints
• Material Properties
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
4 of 9
2. Methodology
1. Process mapping
2. Verification of monitoring tools
3. Program selection
4. Selection of methods
5. Database elaboration
6. Modelling• Block Characteristics• Welding Characteristics from 2D Model• Welding Characteristics from 3D Model• Collection of the historical data
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
General Overview
5 of 9
3. Model
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
Best FittingReal VariationPolynomial Regression – Eight Variables – All FeaturesPolynomial Regression – Two Variables – All FeaturesPolynomial Regression – Two Variables – One FeatureNeural Network – Eight Variables – All FeaturesNeural Network – Two Variables – All Features – 1 HL, 60 HN – 2 STDNeural Network – Two Variables – One Feature – 3 HL, 71 HN – IQRBack Feature SelectionBack Feature Selection - Neural Network – 3 HL, 81 HN - IQR
6 of 9
4. Results and Analysis
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
7 of 9
5. Conclusions
Geometric Parameters
Data Treatment – 4 months
KNIME – straight-forward
IQR – Best with Neural Network
2 STD – Best with Polynomial Regression
Best Fitting (R2) > Polynomial Regression (R2) > Neural Network (R2)
Polynomial Regression with many features – No guess
Reduction of Variables was elementary
More data necessary in order to improve Learning
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
8 of 9
6. Future Works
More data in order to improve the study
How to integrate the solution with current design software
Assess the excess of material (Production Allowance)
Extend study to other structural elements
Marcio Fleming, 7 th EMship cycle: 2016 − 2018 Defence of Master Thesis, La Spezia, February 2018
Accuracy Control and Welding Distortion Prediction in a Deck Plate
9 of 9