AI-enabled compact gas flow meter
Ai-enabled compact gas flow meter
Starting situation
A leading gas flow and pressure meter manufacturer aimed to tap into new market segments with a cost-effective compact device. They trusted IMT, given our historical partnership and expertise in gas flow measurement.
Project objectives
Create a lightweight, compact gas flow meter that offers precision at a production cost 50% less than its competitors. This device needed to cater to diverse uses, from hospital equipment calibration to engineering and field service gas flow assessments.
Technical implementation
The innovative gas flow meter combines space- and cost-efficient design with the highest precision by using two differential pressure sensors and special grids that minimize impurities. AI integration efficiently calculates measurement correction factors, reducing memory requirements by 99% and extending battery life.
The device features an LCD display, and Li-Ion battery, and runs on a powerful in-house operating system for optimal real-time performance. Versatile connectivity options such as USB-C and Bluetooth LE are integrated, and the user-friendly design is optimized for quick installation and optimal visibility.
Insights into the development journey
- Innovative design:
Achieving accuracy in a space and cost-effective model was a hurdle. Our unique, patent-pending mechanical design, featuring two differential pressure sensors divided by grids, rose to the challenge. These grids have larger holes to minimize dust pollution risks, ensuring the meter's long-term reliability across different environments.
- AI-integration:
Traditionally, measurement correction factors were stored in a memory-consuming 4-dimensional look-up table. Instead, we harnessed machine learning using available test measurement data, training a neural network algorithm that predicts these factors. The training was done on a PC, while the neural network was deployed on the embedded hardware of the flow meter. This AI approach requires only 10kB of memory – a massive 99% reduction of memory space. The outcome? Reduced power consumption and extended battery life.
- Connectivity and design:
The device boasts versatile connection options, including USB-C and BLE for mobile apps on both iOS and Android. Recognizing users might place the device on irregular surfaces and need clear screen visibility from multiple angles, we leaned on our UX and ID expertise. After creating multiple 3D prototypes and gathering initial user feedback, we finalized a design that's both functional, user-friendly, and optimized for quick assembly.
- Power and performance:
We incorporated an LCD display and a Li-Ion battery. The software operates on two processors: one for the user interface and power management, the other for tasks like measurement, data processing, and AI inference. Our in-house 'Dataflow Runtime' OS ensures optimal real-time performance and smooth data processing. Dataflow Designer was invaluable for system and software architecture design, thanks to its automatic code generation and documentation features. More details on https://www.imt.ch/en/imt-dataflow.
Results
After a two-year development journey, our new flowmeter hit the market. It promises 1.7% accuracy in flow measurements, a user-friendly interface, and extended battery life. The device is CE-marked and registered as a Class I medical device in the USA.
