AI-based Digitizer Technology Optimises Pipeline Monitoring Tasks

There are many applications that can take advantage of the gains available from the latest intelligent digitizer solutions. These take the form of a high-end AI-based data acquisition (DAQ) card, with support from a rugged and advanced industrial PC (IPC) platform and an appropriately specified graphics processing unit (GPU).

Among the most notable beneficiaries of this technology is set to be the distributed acoustic sensing (DAS) or distributed optical fibre sensing (DOFS) market, where pipelines that extend over many kilometres require acoustic (fibre-optic) sensing for inspection tasks.

Here, a high-end IPC + DAQ solution from Advantech and its co-creation partners can deliver optimal leak detection, flow monitoring and intrusion detection through continuous long-term data acquisition.

According to market research company MarketandMarkets, the DAS/DOFS sector will grow from $462m to $792m between 2020 and 2025, representing a notable CAGR of 11.4%. Critical to this growth, however, is efficient and effective DAQ, a task that is far from simple when it comes to the data-prolific ultrasonic inspection of oil pipelines, for example.

Vast data stream

A DAS system serving an oil pipeline will generate in the region of around 1TB of data per day. Some form of edge computing solution is therefore required to analyse and process this vast quantity of data.

Another challenge is the clear need for AI. Efficient algorithms for the detection of events of interest are of the utmost importance. To explain further, pipeline owners or managers may wish to know more details and classify each type of data ‘event’ detected.

Such events might include someone jumping on a pipeline, or vibration from a flow change or a nearby passing truck. AI can ‘judge’ and classify each event to reduce false positives.

Also high on the list of challenges is interoperability. DAS systems require connection to other systems and the cloud, so they can immediately alert operators and show the event location. This capability allows the monitoring team to take actions such as changing the operation mode of machinery, modifying pipeline parameters or adjusting alarm security.

System components

So what does a solution look like that is able to overcome all of these challenges and deliver the efficient processing of distributed acoustic sensing data using a deep learning approach?

Well, system components will typically consist of: a data acquisition and processing unit to continuously capture large amounts of data; real-time operating systems to analyse the data; and servers in the cloud or on the edge for data archiving, analytics and visualization. 

Although the digitizer market is quite wide in generic terms, Advantech’s latest DAQ card, the PCIE-1840, focuses on common industrial applications that include DAS/DOFS. Offering up to 125MS/s sampling rate to retrieve back scattered laser or ultrasonic signals, a 16-bit resolution digitizer with four simultaneous analogue inputs and 2GB on-board memory offers non-stop, high-definition data streaming capability.

Based on this new product, Advantech can offer different AI-enabled solutions. One comprises of an Advantech IPC-242 V3 ultra-compact IPC with the PCIE-1840 and Quadro RTX A2000 GPU card from the co-creation partner NVIDIA.

This mainstream AI digitizer features small form-factor, four 125MHz/16-bit sensor interfaces, Intel 12^th Gen Core i CPU of 16C/24T and NVIDIA GPU of 3,328 CUDA cores, 8.0 TFLOPs FP32 computing power. It is suitable for small to medium scale applications such as tunnel monitoring or traffic management.

A more advanced configuration sees an Advantech MIC-770W V3 modular IPC with MIC-75G20 GPU expansion module in which PCIE-1840 and Quadro RTX A4500 GPU cards are installed. This high-end AI digitizer features four 125MHz/16-bit sensor interfaces, Intel 12^th Gen Core i CPU of 16C/24T and NVIDIA GPU of 7,168 CUDA cores, 23.7 TFLOPs FP32 computing power.

It is suitable for large scale applications such as oil, electric pipeline monitoring or boarder security surveillance. It can act an edge server which archives data, retrains/deploys AI models and manages digitizers on the field.

Both aforementioned platforms permit the user to add various I/O interfaces in line with project requirements. Such as industrial protocol communication ports or Wi-Fi, LTE to link the digitizer with controllers, machinery or in-house IT systems.

Application topology

In the case of a DAS system serving tunnels, an application-specific topology might apply. For example, the system integrator deploys several units of IPC-242 with Quadro A2000 GPU card on the edge, they therefore become responsible for collecting data from the acoustic/ultrasound sensors before processing and transferring it to the metadata.

From here, data can transfer to a MIC-770 IPC with the high-end A4500 GPU card in the back end. Ultimately, the data can reside in an on-premise cloud for subsequent analysis or be retained locally for AI model training.

Such a topology means the user benefits from a cloud/edge architecture: the private cloud based on the high-end IPC with high-end GPU card; and the edge platforms based on the entry-level IPC with entry-level GPU card. In short, the edge platforms collect data and use the AI model to deliver real time data analysis and respond to local event.

The collected data is also extracted and uploaded to the cloud for archiving and research, the cloud also retrains AI model and deploys it to multiple edge platforms automatically. There are many advantages to this configuration. For example, if an earthquake happens and generates landslide near a tunnel, the use of a high-performance edge platform means the user can detect abnormal acoustic waveform in real time.

In turn, the system can send an alarm to the traffic management authority or tunnel administrator – also in real time. The edge server can monitor the digitizers on the field, also retrains the AI model and automatically deploys it. This saves the SI on-site maintenance cost, whilst the SI can also access the data archived in the edge server for research purpose.

Further use cases

As well as DAS applications such as the ultrasonic inspection and monitoring of pipelines, Advantech IPC + DAQ solutions can benefit many other industries, including automotive. Here, the use of ultrasonics in the non-destructive testing (NDT) of engine parts is widespread at automotive plants across Europe.

Another common application is the inspection of industrial motors using transient waveform capture, where the use of Advantech digitizer technology can help testing departments to perform this task automatically.

Put simply, any application that can gain from proven synchronisation and timing to process sensing data effectively should consider what Advantech can offer with its AI-based IPC + DAQ solutions. Thanks to the company’s renowned co-creation strategy, these ready-to-go solutions can save considerable amounts of development and commissioning time to bring projects onstream efficiently and cost-effectively.

For further information, please contact;

Daniel Sun
daniel.sun@advantech.de
www.advantech.com/en-eu