New Tech Utilizing AI Could Help Mitigate Underwater Noise Pollution
Researchers from the University of Glasgow and the University of British Columbia have developed a groundbreaking system that harnesses the power of artificial intelligence (AI) to model how sound waves travel underwater. This technology has the potential to significantly reduce the ecological impact of noise pollution on marine life. By accurately understanding the movement and spread of sound waves from human activities, industries such as shipping and renewables can make better-informed decisions to minimize their impact on the undersea environment.
The detrimental effects of underwater noise pollution caused by human technologies, such as the propellers of cargo ships and the construction and operation of offshore wind farms, are well-documented. These loud sounds disrupt the migration patterns of marine mammals like dolphins and whales and hinder their ability to navigate using echolocation. To combat this issue, it is crucial to develop a comprehensive understanding of how sound waves from human activities propagate underwater.
Currently, accurately modeling the physics of sound wave movements and transmission loss underwater is a painstakingly difficult process that requires substantial computational power. Large-scale projects can take several days just to model the spread of sound waves through water. To address this computational challenge, researchers turned to deep neural networks and AI to provide real-time feedback on the propagation of sound waves.
In a recently published paper titled Deep neural network for learning wave scattering and interference of underwater acoustics, the researchers explain their acoustic wave modeling system built using a convolutional recurrent autoencoder network (CRAN). This advanced AI model compresses complex modeling data into a simpler form and uses a long short-term memory network to analyze and predict how underwater sound waves propagate over time based on previously learned underwater physics.
The system was trained using 30 different two-dimensional simulations of underwater environments, each with unique seafloor surfaces and sound frequencies. Following extensive training, the CRAN was able to accurately predict the behavior of sound waves in 15 new underwater scenarios, even when encountering wave interactions and scattering by rigid surfaces. The accuracy of the model was impressive, with less than 10% error for a duration more than five times longer than the training data.
Dr. Wrik Mallik from the University of Glasgow’s James Watt School of Engineering, the corresponding author of the paper, expressed his optimism about the results. He highlighted the potential of deep neural networks in predicting the complex physics of underwater ocean acoustic propagation. The ability to produce models of underwater acoustic scattering within seconds, as opposed to days, would be a significant breakthrough for the field. Real-time feedback through devices used on the ocean could lead to more effective planning and help mitigate the effects of noise pollution on marine animals.
While this study is in its early stages, the researchers believe that the CRAN technology can be scaled up to handle fully three-dimensional acoustic simulations. They have already begun further development and refinement of the system, with plans to test it in real-world situations in the coming months.
The potential of AI-driven modeling systems like CRAN in reducing the ecological impact of underwater noise pollution is immense. With its ability to provide rapid and accurate predictions, this technology could revolutionize industries by enabling them to make informed decisions that minimize their effects on the undersea world. By addressing the challenges posed by noise pollution, we can take significant steps toward protecting marine life and preserving the delicate balance of our ocean ecosystems.
