Hyundai Mobis is developing an artificial intelligence system that utilizes sound and applying it to its production line. The AI assesses quality accuracy by listening to subtle sounds generated during the product inspection process.
On the 19th, Hyundai Mobis announced that it has established a pilot Acoustic AI-based inspection system at its Changwon plant, which produces motor-controlled power steering systems.
Acoustic AI is a next-generation AI technology that uses sound as its medium. It differs from the recent surge of generative AI that asks questions and provides answers based on language. There are currently few applicable cases in the manufacturing sector.
The core of Acoustic AI is the development of algorithms that assign meaning to specific sounds and aid in making appropriate judgments. Over the past few years, Hyundai Mobis has devoted efforts to developing various new technologies using AI, showcasing production techniques specialized in quality management.
Starting with the Changwon plant, Hyundai Mobis plans to expand the Acoustic AI inspection system to other parts of the production process. Parts that inevitably generate noise due to movement, such as braking devices, will be prioritized.
Choi Nak-hyun, head of Hyundai Mobis’s DT Promotion Division, stated, “This is a leading case of process innovation not only in the automotive field but in the global manufacturing industry as a whole,” adding that “We will present differentiated new technologies using AI not only in production technology but also in research and development and improving work environments.”
If generative AI emphasizes versatility for the general public, Acoustic AI is suitable for industrial applications in smart factories. Its biggest advantage is the ability to handle large volumes of work in a short time. In fact, the inspection system established at the Changwon plant can inspect the quality of one unit per second.
Currently, the Changwon plant produces 1.3 million units of MDPS (Motor Driven Power Steering) annually. The process consists of 23 steps, including component assembly, vibration, and noise inspection.
Since MDPS is a safety component that allows direct perception of steering performance through the steering wheel, meticulous quality checks are particularly necessary. Among these, noise inspection is conducted with actual power connected to the MDPS. The sound generated by the rotating motor attached to the MDPS displays a consistent wave pattern, and the AI evaluates any deviations or anomalies in the waveform.
Previously, after the assembly process, an automated system would filter out suspected defective products that fell below a baseline, which would then be re-evaluated by trained personnel. To enhance quality competitiveness, Hyundai Mobis successfully developed Acoustic AI after about a year of data collection and analysis with field engineers, production technology experts, and PhD-level AI specialists. This groundbreaking system can even identify the cause of quality issues when they arise.
Meanwhile, Hyundai Mobis plans to confirm the effectiveness of Acoustic AI and apply it more broadly. In the manufacturing field, it aims to develop a proprietary AI model through the meaningful data continuously secured by leading the development and introduction of Acoustic AI. The accuracy of the trained AI is also expected to improve further.
Lee Sang-jin daedusj@autodiary.kr