PRESS-NEWS.org - Press Release Distribution
PRESS RELEASES DISTRIBUTION

KAIST develops AI that automatically detects defects in smart factory manufacturing processes even when conditions change​

2025-08-26
(Press-News.org) Recently, defect detection systems using artificial intelligence (AI) sensor data have been installed in smart factory manufacturing sites. However, when the manufacturing process changes due to machine replacement or variations in temperature, pressure, or speed, existing AI models fail to properly understand the new situation and their performance drops sharply. KAIST researchers have developed AI technology that can accurately detect defects even in such situations without retraining, achieving performance improvements up to 9.42%. This achievement is expected to contribute to reducing AI operating costs and expanding applicability in various fields such as smart factories, healthcare devices, and smart cities.

KAIST (President Kwang Hyung Lee) announced on the 26th of August that a research team led by Professor Jae-Gil Lee from the School of Computing has developed a new “time-series domain adaptation” technology that allows existing AI models to be utilized without additional defect labeling, even when manufacturing processes or equipment change.

Time-series domain adaptation technology enables AI models that handle time-varying data (e.g., temperature changes, machine vibrations, power usage, sensor signals) to maintain stable performance without additional training, even when the training environment (domain) and the actual application environment differ.

Professor Lee’s team paid attention to the fact that the core problem of AI models becoming confused by environmental (domain) changes lies not only in differences in data distribution but also in changes in defect occurrence patterns (label distribution) themselves. For example, in semiconductor wafer processes, the ratio of ring-shaped defects and scratch defects may change due to equipment modifications.

The research team developed a method for decomposing new process sensor data into three components—trends, non-trends, and frequencies—to analyze their characteristics individually. Just as humans detect anomalies by combining pitch, vibration patterns, and periodic changes in machine sounds, AI was enabled to analyze data from multiple perspectives.

In other words, the team developed TA4LS (Time-series domain Adaptation for mitigating Label Shifts) technology, which applies a method of automatically correcting predictions by comparing the results predicted by the existing model with the clustering information of the new process data. Through this, predictions biased toward the defect occurrence patterns of the existing process can be precisely adjusted to match the new process.

In particular, this technology is highly practical because it can be easily combined like an additional plug-in module inserted into existing AI systems without requiring separate complex development. That is, regardless of the AI technology currently being used, it can be applied immediately with only simple additional procedures.

In experiments using four benchmark datasets of time-series domain adaptation (i.e., four types of sensor data in which changes had occurred), the research team achieved up to 9.42% improvement in accuracy compared to existing methods.[TT1]

Especially when process changes caused large differences in label distribution (e.g., defect occurrence patterns), the AI demonstrated remarkable performance improvement by autonomously correcting and distinguishing such differences. These results proved that the technology can be used more effectively without defects in environments that produce small batches of various products, one of the main advantages of smart factories.

Professor Jae-Gil Lee, who supervised the research, said, “This technology solves the retraining problem, which has been the biggest obstacle to the introduction of artificial intelligence in manufacturing. Once commercialized, it will greatly contribute to the spread of smart factories by reducing maintenance costs and improving defect detection rates.”

This research was carried out with Jihye Na, a Ph.D. student at KAIST, as the first author, with Youngeun Nam, a Ph.D. student, and Junhyeok Kang, a researcher at LG AI Research, as co-authors. The research results were presented in August 2025 at KDD (the ACM SIGKDD Conference on Knowledge Discovery and Data Mining), the world’s top academic conference in artificial intelligence and data.
※Paper Title: “Mitigating Source Label Dependency in Time-Series Domain Adaptation under Label Shifts”
※DOI: https://doi.org/10.1145/3711896.3737050

This technology was developed as part of the research outcome of the SW Computing Industry Original Technology Development Program’s SW StarLab project (RS-2020-II200862, DB4DL: Development of Highly Available and High-Performance Distributed In-Memory DBMS for Deep Learning), supported by the Ministry of Science and ICT and the Institute for Information & Communications Technology Planning & Evaluation (IITP).

 

END


ELSE PRESS RELEASES FROM THIS DATE:

Research alert: Alcohol opens the floodgates for bad bacteria

2025-08-26
Alcohol-associated liver disease (ALD) is a major cause of liver transplantation and death worldwide, and its impact is only growing. In 2022, the annual cost of ALD in the United States was $31 billion. By 2040, this number could be as high as $66 billion. ALD has limited therapeutic options, so scientists are looking for new ways to target the molecular biology of ALD to help prevent its occurrence or reduce its severity. Now, scientists at University of California San Diego School of Medicine have found that chronic alcohol use impairs the production of a key cellular signaling ...

American Gastroenterological Association, Latica partner to assess living guidelines using real-world evidence

2025-08-26
Palo Alto, CA and Bethesda, MD (Aug. 26, 2025) – American Gastroenterological Association (AGA) is proud to announce a new collaboration with, Latica, a leader in real-world clinical data and artificial intelligence-driven research solutions, to support a real-world evidence initiative: the IBD Living Guidelines Real-World Evidence Study. The study will be powered by Latica’s proprietary data platform and robust clinical datasets, in partnership with Latica’s GI Network, starting with data from Gastro Health and Allied Digestive Health, two of the largest and most respected community gastroenterology practices ...

University of Tennessee collaborates on NSF grants to improve outcomes through AI

2025-08-26
Faculty members from the Min H. Kao Department of Electrical Engineering and Computer Science at University of Tennessee are involved in two collaborative National Science Foundation grants that aim to address health disparities research and enhance the performance and productivity of AI science. Tabitha Samuel, the interim director and operations group leader for UT’s National Institute for Computational Sciences (NICS), is the principal investigator for UT on both projects. AI Advancement in Health ...

New technique at HonorHealth Research Institute uses ultrasound to activate drugs targeting pancreatic cancer

2025-08-26
SCOTTSDALE, Ariz. — Aug. 16, 2025 — In a quest for ever-more-effective treatments for pancreatic cancer, HonorHealth Research Institute is combining the power of targeted drugs and ultra-sound technology to shrink the tumors of patients whose cancer has not spread to other parts of the body but is too advanced to allow safe surgical removal. Acoustic Cluster Therapy (ACT) uses tiny clusters of gas bubbles and oil droplets that make up PS101, which is combined with a modified version of FOLFIRINOX, an FDA-approved combination of four ...

Companies 'dumbed down' cryptocurrency disclosures in good markets prior to reporting standardization, Rotman research finds

2025-08-26
August 26, 2025 Companies 'dumbed down' cryptocurrency disclosures in good markets prior to reporting standardization, Rotman research finds Toronto - In the run-up to the introduction of accounting standards for the reporting of cryptocurrency, companies took various approaches to disclosing their involvement with the novel exchange medium. Investors had to rely more on contextual narrative explanations about a company's crypto involvement instead of getting it from numerical data. An analysis of reporting ...

MSU study: What defines a life well-lived? Obituaries may have the answers.

2025-08-26
Why this matters: Obituaries function as time capsules that reflect what people, at any given moment in history, understand to be a life well-lived. By studying how obituaries evolve across time and context, we can gain deeper insight into how societies define a worthy life, express loss and pass on values across generations. New MSU-led research takes a novel approach to the psychological study of legacy by focusing on how individuals are actually remembered by others instead of how they wish to be remembered. Legacy motivations influence a range of real-world ...

Wind isn’t the only threat: USF-led scientists urge shift to more informed hurricane scale

2025-08-26
Key takeaways: Storm surge and rainfall — not wind — cause the majority of hurricane deaths, yet are absent from the current warning system. Researchers developed and tested the Tropical Cyclone Severity Scale, which incorporates all three hazards and can rate storms up to Category 6. Study shows people are more likely to correctly identify risks and evacuate when informed using the new scale. TAMPA, Fla. (Aug. 26, 2025) – Wind alone does not account for all hurricane-related fatalities. Storm surge and rainfall do as well. Yet the current warning system – the Saffir-Simpson Hurricane ...

Study: Fossils reveal reliable record of marine ecosystem functioning

2025-08-26
A new study published in the Proceedings of the National Academy of Sciences confirms that fossilized remains of marine invertebrates can accurately reflect the functional diversity of past ecosystems—offering a powerful tool for understanding long-term ecological change and informing modern conservation. UNLV geoscience professor Carrie Tyler, in collaboration with Michał Kowalewski from the University of Florida, compared living marine communities with their corresponding skeletal remains and the fossil record across 51 coastal sites in ...

New Simon Fraser University–University of Exeter partnership fast-tracks path to become a lawyer

2025-08-26
Simon Fraser University is breaking new ground in legal education with an exclusive international partnership that offers students a faster route to becoming a lawyer. The new SFU-Exeter Accelerated Law Program gives students the opportunity to graduate in only six years with a Bachelor of Arts (BA) from SFU and a Bachelor of Laws (LLB) from the University of Exeter, qualifying them to pursue legal careers in Canada and the UK. The University of Exeter is a member of the elite Russell Group, a collection of the UK’s most distinguished research-intensive universities. ...

Busy bees can build the right hive from tricky foundations

2025-08-26
There’s more than one way to build a honeybee hive, depending on the needs of the bees, according to a study published August 26th in the open-access journal PLOS Biology by Golnar Gharooni-Fard of the University of Colorado Boulder, USA, and colleagues. Honeybees are renowned for their ability to build intricate hives where they can store their food and raise their larvae. Hive construction is the collaborative effort of thousands of hard-working bees, which also demonstrate a knack for adjusting their honeycomb structures to account for available space and resources. In this study, Gharooni-Fard and colleagues ...

LAST 30 PRESS RELEASES:

Engineers develop solid lubricant to replace toxic materials in farming

Repurposing gemstone polishing waste to create smart cement

Patient-physician messaging by race, ethnicity, insurance type, and preferred language

Unrecognized motor difficulties and developmental coordination disorder in preschool children

Background genetic variants influence clinical features in complex disorders

Smarter battery tech knows whether your EV will make it home

Overactive microRNAs block fat cell development in progeria

Crosswalk confusion: MA drivers flummoxed by pedestrian hybrid beacons, find UMass Amherst researchers

Study shows heart disease mortality disproportionately burdens low-income communities in California

Intracardiac echocardiography recognized as ‘transformative’ imaging modality in new SCAI position statement

Study finds ‘man’s best friend’ slows cellular aging in female veterans

To get representative health data, researchers hand out fitbits

Hiring in high-growth firms: new study explores the timing of organizational changes

Boosting work engagement through a simple smartphone diary

Climate change may create ‘ecological trap’ for species who can’t adapt

Scientists create ChatGPT-like AI model for neuroscience to build one of the most detailed mouse brain maps to date

AI and omics unlock personalized drugs and RNA therapies for heart disease

2023 ocean heatwave ‘unprecedented but not unexpected’

Johns Hopkins researchers develop AI to predict risk of US car crashes

New drug combination offers hope for men with advanced prostate cancer

New discovery finds gene converts insulin-producing cells into blood-sugar boosters

Powerful and precise multi-color lasers now fit on a single chip

Scientists agree chemicals can affect behavior, but industry workers more reluctant about safety testing

DNA nanospring measures cellular motor power

Elsevier Foundation and RIKEN launch “Envisioning Futures” report: paving the way for gender equity and women’s leadership in Japanese research

Researchers discover enlarged areas of the spinal cord in fish, previously found only in four-limbed vertebrates

Bipolar disorder heterogeneity decoded: transforming global psychiatric treatment approaches

Catching Alport syndrome through universal age-3 urine screening

Instructions help you remember something better than emotions or a good night’s sleep

Solar energy is now the world’s cheapest source of power, a Surrey study finds

[Press-News.org] KAIST develops AI that automatically detects defects in smart factory manufacturing processes even when conditions change​