Motivation, Background, and Objective:
Photoacoustic imaging is a non-ionizing imaging technique that provides functional and structural information about imaging targets with optical contrast and ultrasound resolution. Thus, it is widely utilized in the medical field. However, in recent years, photoacoustic imaging techniques have been used in industrial applications as non-distractive testing such as surface crack detection. However, the photoacoustic system cost and the time consumption for scanning and imaging reconstruction limit using it in non-destructive testing. Therefore, in this study, low-cost photoacoustic equipment with machine learning technique will be investigated in surface crack detection. This investigation reduces the system complexity and the time consumption for detection compared with that when photoacoustic imaging techniques are used.
Statement of Contribution/Methods:
A piece of black plastic with four surface cracks in a water tank was used as a phantom. The width of each crack was almost 1mm. The photoacoustic emissions were generated from this phantom using a pulse laser diode (PLD) (905D5S3J08X). The width of each firing laser pulse was 100 nm. In addition, the output optical energy and wavelength of this PLD were 3 micro-J and 905 nm respectively. In this experiment, photoacoustic emissions were acquired by using an open-source ultrasound board with one element ultrasound transducer (C310-SU). The centre frequency and bandwidth of this transducer were 5MHz and 90% of the centre frequency respectively. The scanning step of the ultrasound transducer in the lateral direction was 0.1 mm. The received photoacoustic emission for each scanning point was averaged 10 times before using it to improve the SNR of the received photoacoustic signal. In this experiment, 1131 photoacoustic signals were acquired from a cracked place and 4522 photoacoustic signals were acquired from an uncracked place. These received signals were converted to spectrogram images before using them in the machine learning model. In this study, a Convolutional Neural Network (CNN) classification model was used for detecting the presence of cracks in the surfaces. The dataset was randomly split into two independent parts with 80% and 20% for training and testing, respectively.
Results, Discussion, and Conclusions:
The offline prediction accuracy of this model was 97%. This accuracy makes the system capable of performing real-time detection for the crack. When compare this machine learning detection technique with the imaging technique, the machine learning technique reduces the complexity of the system. This is because one scanning line is used as input to the machine learning model, unlike the imaging technique that needs multiple scanning lines to reconstruct the photoacoustic image. The acquisition of multiple scanning lines is time consuming if one single element ultrasound transducer is used. In addition, the cost of the imaging system will be expensive if a multiple element ultrasound transducer is used. Moreover, when a machine learning technique is used, different levels of cracks can be detected. This is unlike the imaging technique that is affected by the contrast level of imaging targets.
