“Regular inspection of nuclear power plant components is important to guarantee safe operations,” said Mohammad R. Jahanshahi, an assistant professor in Purdue’s Lyles School of Civil Engineering. “However, current practice is time-consuming, tedious, and subjective and involves human technicians reviewing inspection videos to identify cracks on reactors.”
Making the inspection process more difficult, nuclear reactors are submerged in water to maintain cooling, and this makes any inspection of the reactor’s component parts unfeasible due to the high temperatures and radiation levels. Currently, inspections are made using remotely controlled videos of the underwater reactor surfaces, and this can lead to human error.
Cracks can lead to leaks
The U.S. is the world’s leading supplier of commercial nuclear power, providing almost 20 percent of the nation’s total electrical generation. However, between 1952 to 2010, there have been 99 major nuclear power incidents worldwide that cost more than $20 billion and led to 4,000 fatalities. A total of 56 out of the 99 incidents occurred in the U.S.
“One important factor behind these incidents has been cracking that can lead to leaking,” Jahanshahi says. “Nineteen of the above incidents were related to cracking or leaking, costing $2 billion. Aging degradation is the main cause that leads to function losses and safety impairments caused by cracking, fatigue, embrittlement, wear, erosion, corrosion, and oxidation.”
He goes on to explain there are a number of Vision-backed methods used for detection of cracks on concrete, rock, or pavement surfaces, but little attention has been given to detecting cracks on metallic surfaces. The few methods that do exist rely on a single image.
“If a crack is not detected in these single images or a noisy pattern is falsely detected as a crack in the image, no other information is available to correct the detection results,” he says.
The naïve Bayes-convolutional neural network
Purdue researchers have a system under development called the naïve Bayes-convolutional neural network (NB_CNN). It’s a “deep learning” framework that analyses individual video frames for crack detection. Using an innovative “data fusion scheme,” information from each video frame is aggregated to enhance the overall performance of the system.
Convolutional neural networks are a form of artificial intelligence called “deep learning,” and is used in a number of commercial applications such as facial and speech recognition platforms. Perdue’s CNN system detects cracks in overlapping patches in each video frame. Then, using the data fusion algorithm scheme, or NB, the crack can be detected from one frame to the next, while discarding false-positives. Using this method, researchers have been able to achieve a 98.3 percent success rate,
The algorithm “mimics the ability of human vision to scrutinize cracks from different angles, which is important because some cracks are obscured by the play of light and shadow,” Jahanshahi says. “Data fusion comes up with more robust decision making than would otherwise be possible,” he adds.
The findings are detailed in a paper published in October in the journal IEEE Transactions on Industrial Electronics. The paper is entitled NB-CNN: Deep Learning-based Crack Detection Using Convolutional Neural Network and Naïve Bayes Data Fusion.