Preventing the failure of large-scale facilities such as nuclear power plants! The SIAT analytics technology that analyzes vast quantities of sensor data to find anomalies
- AsakuraThere are a lot of possibilities. For example, optimization of the maintenance and inspection process. Until now, maintenance staff at places like factories and power plants have devoted an enormous amount of time and effort to inspections, tapping each section to examine it or performing visual checks based on their experience. SIAT can streamline this process.
- MitaHow does that work?
- AsakuraYou make models at regular intervals and compare them. Then, when there are issues such as frictional wear or degradation, the impact of this will start show around that area. In other words, changes will appear in the model. This enables you to detect when a failure is likely to occur.
- YoshihiraThe detection of when a failure is likely to a occur is a key point. You can only stop operation to perform maintenance and inspection for a limited time, and if you try to replace or inspect all components, you won't be able to keep up. For this reason, it's better to give priority to checking points that appear likely to fail. Then you don't have to replace parts that don't need replacing, operations are suspended for a shorter time, and you can raise the utilization rate while cutting costs at the same time.
- MitaInspecting points that are likely to fail at large facilities would certainly be more efficient!
- AsakuraIt can also be utilized for quality control. For example, at an ironworks a range of sensors are used to track heat and pressure during the iron production process. If you then build a model based on when good quality iron was produced, you can control specific aspects to maintain quality, such as raising the temperature when it drops too low.
- MitaSo SIAT can help us understand things that we once relied on the expertise of a craftsman for?
- YoshihiraNot all expertise can be modeled, but there is plenty of potential in that area. Even if the generation of craftsman active today disappear without passing on their techniques, you may be able to preserve them by creating a model using SIAT.
- AsakuraI also believe this technology is helpful for keeping tabs on the margin of safety. Inputting a value for a certain sensor on a model SIAT created will affect the sensors related to it, and be propagated down the line to all other connected sensors. Using this means you can simulate how severe an anomaly the system can handle safely.
- YoshihiraThis has already been applied to the world of computers(*). For example, imagine that the load on a Web server spikes suddenly. When this happens, you can simulate whether the response speed of the database server is acceptable. With previous techniques it was difficult to raise the accuracy of a simulation, but because SIAT bases relationships on actual legacy data, it is possible to perform more accurate and qualitative evaluations. If a database server is unable to withstand the load, you can take qualitative steps to produce a design that copes with this, such as adding another server. We are trying to apply this to structures such as factories and power plants as well.
(* NEC offers a "WebSAM Invariant Analyzer" product that utilizes invariant analysis technology for the operation and management of ICT systems)
- MitaIf you don't mind me asking, what sort of setups are SIAT compatible with?
- AsakuraAs long as it involves sensors, SIAT can be used with basically anything. Aside from factories and power plants, there is also a high probability of it being applicable to structures such as bridges and buildings that have sensors.
- MitaIt sounds like it can be utilized in a myriad of situations. I was worried about older buildings collapsing if a large earthquake hit, but using SIAT enables you to fix any points that appear to be failing before disaster strikes! The sheer potential of SIAT makes me feel giddy with excitement.
- YoshihiraIncidentally, with the cooperation of Chugoku Electric Power Company, NEC carried out proof-of-concept tests at the technical training facility for their Shimane Nuclear Power Plant to find points that could fail in the future, but weren't detectable using previous techniques. We obtained very pleasing results, such as being able to detect problems around 20 times faster than conventional techniques or the human eye.
- MitaThat sounds encouraging! Are other manufacturers developing technology like this?
- YoshihiraNo. There hasn't been any other technology that has taken this unique an approach. Although we aren't doing anything that difficult from an analytical technology perspective, as mentioned earlier, the concept of getting an overall picture of system behavior was in itself very innovative. At the time, SIAT technology was actually so unique that many people in the lab were skeptical. But although this was a new and unique technology, everyone showed a lot of interest in SIAT in the NEC operations department, which takes a proactive approach to finding practical applications. There was some negative feedback in the beginning, but we were only able to implement this technology through the help we received from the operations department, such as evaluation of actual systems. I'm lucky to work as a researcher at a company with people like this.
- MitaIt seems like this kind of climate at NEC also played a part in making SIAT a reality. I hope to see it used for all sorts of applications going forward. Thank you for sharing your precious time with us today!
- Yoshihira, AsakuraIt was a pleasure.
In this 16th installment, we introduced the SIAT analysis technology that detects anomalies and finds signs of faults at factories and power plants by analyzing vast quantities of sensor data. Did you find it interesting? I was surprised that it could be used to find indicators of future failures, rather than simply things that have already broken. I'm very curious to see how it will be put to use in the coming years.
See you in the next installment of "MiTA TV"!
(Published July 17, 2013)