Neural Network Architectures and Feature Extraction for Lumber Production Prediction

Vincent Martineau,Michael Morin,Philippe Thomas,Jonathan Gaudreault,Hind Bril El-Haouzi

doi:10.21428/594757db.89eadeff

Abstract

We tackle the problem of predicting the lumber products resulting from the break down of the logs at a given sawmill. Although previous studies have shown that supervised learning is well suited for that prediction problem, to our knowledge, there exists only one approach using the 3D log scans as inputs and it is based on the iterative closest-point algorithm. In this paper, we evaluate the combination of neural network architectures (multilayer perceptron, residual network and PointNet) and log representation as input (industry know-how-based features, 2D projections, and 3D point clouds) in the context of lumber production prediction. Our study not only shows that it is possible to predict the output of a sawmill using neural networks, but also that there is value in combining industry know-how-based features and 3D point clouds in various network architectures.

Highlights

The task of forecasting the lumber products resulting from the break down of logs at a given sawmill is complex
We propose to use neural networks which are known to perform well in contexts where a large quantity of information is available
We developed and evaluated 34 combinations of neural network architectures, input representations, output modes and training processes for the lumber production prediction problem

Summary

Introduction

The task of forecasting the lumber products resulting from the break down of logs at a given sawmill is complex. A log break down generally results in multiple lumber products of different value at the same time. A sawmill of moderate size can produce more than eighty different products which could lead to more than a thousand feasible combinations of lumber products to choose from when predicting the outcome of the break down of a single log. Good production forecasts can lead tremendous gains to a forest-product company due to better planning opportunities [1]. When the output of a plant for a given set of logs is known (or estimated) prior to processing, the assignment can be optimized which in turn leads to a better response to the demand and a higher production value

Objectives

Methods

Results

Conclusion