How data acquisition can grow your business
Monday 31 October 2022
While collecting more and more data to process and analyze, the question becomes relevant how to deal with all that data in the most effective way. It can easily take up to 9 months to develop a custom hardware solution that serves that specific purpose if you start from scratch. Fortunately, there are faster solutions to reach that goal, based on design re-use and functional building blocks for board designs, FPGA firmware and (embedded) software infrastructure.
Whether it is a self-driving car or a factory machine: current devices require dealing with complex data in an ever-faster fashion. With nanosecond resolution, they need to collect data, process and interpret them, to make the proper operational decisions. If you can be faster at implementing this, it will give your business a competitive advantage.
Faster and better
Although the trend currently is that the cloud “solves it all”, the solution for these kinds of systems is a different one. It is a combination of data reduction and pre-processing, on-the-edge, and actual signal processing in the cloud; requiring connectivity with sufficient bandwidth. ‘Software and design re-use, such as the use of system-on-modules, can do the trick to acquire, pre-process and communicate data’, says Dirk van den Heuvel, product manager at TOPIC. ‘Uploading these results to the cloud for further algorithmic artificial intelligence (AI) or machine learning (ML) based processing, leads to a quick interpretation and analysis for decision-making.’ And there are even more advantages. Van den Heuvel: ‘The development time of solutions based on this approach typically reduces the development effort by months and it can quite easily be re-used. Another major advantage is the reduction of certification effort that is required for more and more applications.’
A case: ultrasound steel plate inspection
Let’s look at a case of the manufacturing of large steel sheets at a foundry. For quality control, it is essential to check the sheets for any inconsistencies during production using e.g., ultrasound sensors. Such a sheet travels the production process at a speed of 50 km/h. ‘With our solution, we measure the sheets with 96 ultrasonic sensor heads, sampling the sensors at 100MHz and process the acquired data using 24 concurrently operating Miami System-on-Modules. The total raw data rate of over 100Gbps is correlated and processed in concurrence and at real-time using this distributed edge-configuration. The system can execute algorithms that require up to 2T (2^12) multiply-accumulate operations per second or more then 250G FLOPS, supported by 50Mbyte of distributed memory’. By processing the ultrasonic echoes at the sensors, the continuous sheet production rate can be matched. The resulting data stream is reduced to a manageable level and pushed via standard Ethernet connectivity to a (local) cloud server for further processing. How long does the check from sensor measurement to result takes? Just seconds.’
For more information on our solutions check : https://topic.nl/en/expertises