00413234512PH-51 Impulse HammerMeasurement DataWheel separation accidents continue to increase“AI excels at identifying and judging subtle differences that used to depend on human sensory perceptions and knack. By applying AI, we can assist inspection tasks that previously relied solely on human sensory judgments, enabling automated and reproducible assessments. Rion has been manufacturing impulse hammers for vibration analysis for quite some time now. We also have experience incorporating AI into products like hearing aids. So, we started by looking for ways to combine and modify these elements to create a system.“During development, we collected vibration data under various conditions, including different hammering techniques and varying degrees of looseness, to train the AI and build the AI model. We could have used audio data, but that presents the risk of interference from ambient noise. Since we could obtain vibration data with an impulse hammer, we decided to start using vibration instead of sound. Now, we’re at a point where the system automatically assists pass/fail judgment based on the result of hammering if the looseness of the nut is at a clearly detectable level.”The system is nearing practical implementation, but many issues remain to be resolved. The primary challenge is the data acquisition method used to build the AI model.“The first major obstacle was the significant inter-personnel variation in the collected data. Different mechanics strike dif-ferent points, from different angles, and with different force. To calibrate for these differences, we collected data from varying test operators and hammering conditions. We’re also continuing to explore methods for preprocessing the data before use for AI training.“Beyond this, although we’re currently doing experiments using parts from ordinary cars, we need to proceed with an experiment using actual truck parts because the size and num-ber of nuts differ. The nuts used in trucks are significantly larger and heavier than those used for our current experiments, so the waveforms obtained are different and may generate different variability in the data. We’re currently looking for partners who will let us purchase just the actual wheels for large trucks.”Rion refines its technologies, including acoustic, vibration, and particle counting, to identify solutions to social issues. The system they’re currently developing aims to offer solutions for a wide range of social issues beyond truck wheels.“To prevent wheel separation accidents, we will continue to work on enhancing the system’s accuracy before introducing it to the market. Trucks aren’t the only things to use large nuts. All types of equipment would benefit from a system like ours. We’re currently collecting vibration data using an impulse hammer to detect looseness, but we plan to explore using other parameters, like sound and electrical current, to achieve even better accura-cy. Rion has knowledge of not just AI technologies, but hardware technologies like impulse hammers, sound level meters, and accelerometers. Our strength when focusing on a specific chal-lenge is the capacity to address issues from both hardware and software perspectives.”A system for assisting pass/fail judgments that combines an impulse hammer and AIProviding solutions to social issues through both hardware and software, not just AI6756454114014213112311281H26H27H28H29H30H31/R1R2R3R4R50.00240.00250.00261.51.00.50.0023(cases)180Number of incidents (cases)16014012010080604020 (Source: Ministry of Land, Infrastructure, Transport and Tourism, “Occurrence and trend analysis of wheel separation accidents involving large vehicles during 2023”)OKNG0.00270.0028Voltage（V）Time（S）The PH-51 is a hammer with a force sensor built into its head. The sensor detects the vibrations generated when the hammer hits a target, then the vibrations are converted into electrical signals used for analysis on a computer or other equipment. The impulse hammer thus enables evaluation of the vibration characteristics of the target, including its natural frequency.Using AI to detect loose bolts and nutsPlots of the waveforms for Normal (OK) and Abnormal (NG) used in impact inspections. The graph shows data aggregated from 15 measurements. Blue indicates Pass cases; red indicates Fail cases. The horizontal and vertical axes, respectively, represent time (S) and impulse hammer voltage (V). In this way, we can visually confirm the nut loosening.Compiled by the Ministry of Land, Infrastructure, Transport and Tourism, the graph above shows the annual total incidence of wheel separation accidents involving large vehicles. Despite years in which the incidence was lower than in the previous year, the graph shows an overall upward trend. A comparison of the numbers from 2023 (Reiwa 5) with those from 2014 (Heisei 26) shows a near-threefold increase in the number of accidents.Shinichi MasunariAI R&D Group, R&D Department, R&D Center. Joined Rion in 2007. He worked on the design and development of hearing test equipment, including audiometers and auditory evoked potential devices. He was assigned to the AI R&D Group in 2025, where he remains to this day. His hobbies include shogi, which he plays at least once daily against human opponents via an online app.Incidents involving personal injury (cases)(Japanese era year) System for Assisting Pass/Fail Judgments15
元のページ ../index.html#17