ITRI TODAY Fall issue 2018

Gas Leakage Automatic Recognition Technology

Gas leakage has always been a vital safety issue for the chemical engineering and energy industries. To prevent accidents, early detection of gas leakage is crucial; however, it is difficult to conduct long-term monitoring of gas pipelines for minor leakages and it is time- and labor-consuming to complete the inspection for an entire plant. ITRI thus developed Gas Leakage Automatic Recognition Technology (GLART), which can overcome gas leakage detection limits through infrared radiation (IR) image enhancement and auto-recognition via AI technologies. This innovation enables automatic high-speed monitoring of large-area gas pipelines and chemical plants.

Most conventional leakage-monitoring devices that monitor pipeline pressure, flow rate, or even acoustic signals are unable to detect small leaks. In fact, these fixed devices only work well when the leakage rate reaches more than 1% of the total transport amount. GLART, however, can detect leakages as small as 1 gram per hour for methane or other common chemical gases. Moreover, the current data indicates that with ITRI’s technology, the detection rate of minor leakage is increased to 85%-97%, whereas manual inspection using IR camera can only reach an accuracy of 50%.

GLART’s IR image enhancement technology utilizes an image stabilizing compensation method to solve jittering issues in recording and identifies multiple moving features in the images to determine a final motion vector. After that, images are analyzed pixel-by-pixel for differences among timewise consequential images. These differences are then weighted over multiple frames, filtered to exclude abnormality, and treated by various feature-dependent transfer functions. The results are then superposed on the original image. Even minor traces of leakages now become identifiable after image enhancement.

To make gas leakage detection fully automatic, ITRI also developed image auto-recognition technology to replace manual inspection. The R&D team used seven image characteristics that can indicate gas leakage in training machine learning models. With enough learning, a single frame can provide sufficient information for auto recognition of gas leakage in the detection system. This is particularly useful during real-time auto inspection when the camera needs to move very fast.

Combined with existing robots or unmanned aerial vehicles, GLART makes fully automatic pipeline inspection cost effective. Since the software that GLART uses is portable, this technology can also be applied to IR thermal imagers of any brand and model, improving the leakage inspection accuracy and speed for manual operation as well.

In sum, GLART utilizes image processing and machine learning technologies for high-precision detection of minor gas leakage, removing the restriction of traditional IR detection. Therefore, it is capable of performing automatic inspections and thus significantly enhancing the inspection scope and frequency needed in the chemical and energy industries. The deployment of this technology solves the inspection obstacles for elevated pipelines and will drastically improve public and industrial safety worldwide.