Vibration-Based Strength Grading of Sawn Timber Using Piezoceramic Transducers and One-Dimensional Convolutional Neural Networks Osama Abdeljaber, Anders Olsson and Welf Löwe

(1)

Vibration-Based Strength Grading of Sawn Timber Using Piezoceramic Transducers and

One-Dimensional Convolutional Neural Networks

Osama Abdeljaber, Anders Olsson and Welf Löwe

(2)

Timber strength grading: to predict the load-carrying

capacity of timber boards in a nondestructive manner .

(3)

• The current strength grading methods are not very accurate (

• The aim of the proposed seed project: is to develop a

better method for strength grading of timber by making

use of piezoelectric transducers and 1D CNNs.

(4)

The proposed method

Wideband Random Excitation (0-125 kHz) Vibration Signals

Bending Strength

1D CNN

Piezoceramic Transducer

(Sensor)

Piezoceramic Transducer

(Actuator)

(5)

Why Piezoelectric Patches?

Why 1D CNNs?

What will be the outcome of

the proposed seed-project?

(6)

Why piezoelectric patches?

• They can be used for both actuating and sensing.

• They are inexpensive.

• They are capable of exciting boards at very high frequencies.

• They have been successfully used in condition monitoring of

steel and concrete structures.

(7)

Detection of corrosion in reinforced concrete using piezoelectric patches

(Li, 2019)

Detection of cracks in steel members using piezoelectric patches

(Park, 2016)

Why piezoelectric patches?

(8)

Why 1D CNNs?

• 1D CNNs can deal with raw signals without prior preprocessing or feature extraction.

• 1D CNNs merge feature extraction and feature classification into a single learning body.

• 1D CNNs are computationally efficient.

• 1D CNNs have achieved the state-of-the-art performance in challenging

tasks:

(9)

Detection of bolt loosening in steel connections

(Abdeljaber, 2017)

Why 1D CNNs?

Detecting anomalies in electrocardiograph

(ECG) signals. (Kiranyaz et al. 2016).

(10)

What is the outcome of the proposed seed-project?

1. An initial version of the proposed strength grading method.

2. An external funding proposal to be submitted to either FORMAS or KK Foundation.

3. Further collaboration between the department of building technology and DISA on other data-intensive applications:

• Detection of defects in timber from laser-scanned images using generative adversarial neural networks (GANs).

• Machine learning for structural health monitoring of cross laminated timber

structures.

(11)

Vibration-Based Strength Grading of Sawn Timber Using Piezoceramic Transducers and One-Dimensional Convolutional Neural Networks Osama Abdeljaber, Anders Olsson and Welf Löwe

Vibration-Based Strength Grading of Sawn Timber Using Piezoceramic Transducers and

One-Dimensional Convolutional Neural Networks

Osama Abdeljaber, Anders Olsson and Welf Löwe

Timber strength grading: to predict the load-carrying

capacity of timber boards in a nondestructive manner .

• The current strength grading methods are not very accurate (

• The aim of the proposed seed project: is to develop a

better method for strength grading of timber by making

use of piezoelectric transducers and 1D CNNs.

The proposed method

Why Piezoelectric Patches?

Why 1D CNNs?

What will be the outcome of

the proposed seed-project?

Why piezoelectric patches?

• They can be used for both actuating and sensing.

• They are inexpensive.

• They are capable of exciting boards at very high frequencies.

• They have been successfully used in condition monitoring of

steel and concrete structures.

Detection of corrosion in reinforced concrete using piezoelectric patches

(Li, 2019)

Detection of cracks in steel members using piezoelectric patches

(Park, 2016)

Why piezoelectric patches?

Why 1D CNNs?

• 1D CNNs can deal with raw signals without prior preprocessing or feature extraction.

• 1D CNNs merge feature extraction and feature classification into a single learning body.

• 1D CNNs are computationally efficient.

• 1D CNNs have achieved the state-of-the-art performance in challenging

tasks:

Detection of bolt loosening in steel connections

(Abdeljaber, 2017)

Why 1D CNNs?

Detecting anomalies in electrocardiograph

(ECG) signals. (Kiranyaz et al. 2016).

What is the outcome of the proposed seed-project?

1. An initial version of the proposed strength grading method.

2. An external funding proposal to be submitted to either FORMAS or KK Foundation.

3. Further collaboration between the department of building technology and DISA on other data-intensive applications:

• Detection of defects in timber from laser-scanned images using generative adversarial neural networks (GANs).

• Machine learning for structural health monitoring of cross laminated timber

structures.

Questions?