The potential of a weight detection system for forwarders using an artificial neural network

Geiger, Chris; Greff, Daniel; Geimer, Marcus; Starke, Michael; Ziesak, Martin (2018). The potential of a weight detection system for forwarders using an artificial neural network In: FORMEC 2018 – Improved Forest Mechanisation: mobilizing natural resources and preventing wildfires (pp. 157-164). Madrid: Fundación Conde del Vallede Salazar

[img] Text
Proceedings2018 1.PDF - Published Version
Restricted to registered users only
Available under License Publisher holds Copyright.

Download (415kB) | Request a copy

To manage forest processes, information about the amount of timber that passed through different processing steps within the timber supply chain is essential. This information can be used both to keep the overview of the amount of timber already produced for initiating further activities and to clearaccrued operation expenses promptly. When felling operations have been supported by or carried outexclusively through manual working steps, information sources like harvester protocols are either not available or lack sufficient accuracy for their proper use. To solve this problem in cut-to-length logging, a weight-detection system can be integrated into the loading process of the logs as single stems or stem bundles. This system can be used for estimating the timber volume moved, thus closing the information gap without further interactional effort. Unlike common crane scales, which are not available for every type of machine, the method presented here for a weight-detection system is universally applicable in modern crane types. In the first step of the development stage, loading processes during the work task of the forwarder are detected. Using an artificial neural network and data provided by the sensor set of a modern crane, the weight of the load in the grapple can be estimated based on the position of the grapple and the hydraulic pressure of the boom cylinder during these loading processes. The accuracy error of only 10% of the volume moved after initial training of the neural network demonstrates the potential of this system. By obtaining more training data through further measurements, an accuracy error of less than 4% is expected. For decision making within the forwarding process, this technique can therefore be a reasonable solution for short-term and extensive implementation of the smart forwarding processes that are currently being established in the Forwarder2020 project.

Item Type:

Conference or Workshop Item (Paper)

Division/Institute:

School of Agricultural, Forest and Food Sciences HAFL > Multifunctional forest management

Name:

Geiger, Chris;
Greff, Daniel;
Geimer, Marcus;
Starke, Michael and
Ziesak, Martin

Subjects:

S Agriculture > SD Forestry

ISBN:

978‐84‐96442‐84‐9

Publisher:

Fundación Conde del Vallede Salazar

Language:

English

Submitter:

Admin import user

Date Deposited:

18 Feb 2020 08:33

Last Modified:

18 Feb 2020 08:34

Uncontrolled Keywords:

forest management, process management, crane scale, weight detection, neural networks, Forwarder2020

ARBOR DOI:

10.24451/arbor.7345

URI:

https://arbor.bfh.ch/id/eprint/7345

Actions (login required)

View Item View Item
Provide Feedback