If you’re a photographer, you can mess with, for example, masks, burns-and-dodges and local retouching in Photoshop to get landscape horizons/skies and portraits right (including lips, teeth, eyes, blemishes, etc.)—or you can use an AI-based tool like LuminarAI to do the heavy lifting (simply adjust a few sliders) when your workflow means getting projects out the door. Likewise, AI-based industrial tools should keep your processes running transparently.
No question—AI has been overhyped in many ways, at least so far. But when it comes to fine-tuning your process, you don’t have time for hand-coding, compiling and testing. You just want to fill in the blanks, point-and-click to make connections, and maybe work with some templates and/or macros. Like Luminar, AI should be transparent in the tools you use.
Canvass AI has set out to transform manufacturing by demystifying AI in plant operations so non-data scientists can use it. The company empowers manufacturers by providing a no-code platform, allowing operators to evolve their systems from reactive to predictive, which is critical for the future of sustainable and profitable operations.
I spoke with Humera Malik, CEO of the 4-year-old Canvass AI, to find out where AI is taking the company.
Humera Malik, CEO of Canvass AI (Source: Canvass AI)
FE: How did Canvass get started? What were your goals?
Humera Malik: Prior to founding Canvass, I had come out of the world of IoT sensors and infrastructure deployment. It became evident to me that industrials were good at hoarding data but didn’t know how to extract value from it. I found they didn’t know how to access and make use of the large volume and velocity of their data. If they wanted to touch their data and make use of it, they had to bring in outside consultants to do so. I wanted to solve that problem. Hence, Canvass was born.
FE: You support four manufacturing segments: food and beverage, energy, automotive, and metals and chemicals. How did you settle on food?
Malik: For AI to go from hype to reality—we wanted to focus on manufacturing segments where AI can deliver impactful value and where this success could easily be repeated and scaled across different processes and plants. The food processing industry ticks all of these boxes because they have real challenges that need to be urgently addressed. For example, it has been estimated that food manufacturers lose between 16-36% of their raw materials during manufacturing, increasing their overall production costs and creating a food waste issue. What’s more, energy consumption—which is deemed one of the leading costs for food manufacturers—is expected to increase from 11% to 16%. These issues combined show that the food industry can’t afford to keep operating in the status quo.
With AI, operators now have the data processing power and speed to predict costly errors across highly complex and dynamic production processes.
FE: How does Canvass work?
Malik: One of the key differences of Canvass is that it is developed for the industrial operators to use—not teams of data scientists or IT. As a no-code machine learning platform, Canvass puts the power of AI in the hands of the process experts by simplifying the process of building, training and scaling applied industrial AI in their day-to-day operations—without requiring coding skills. By operationalizing machine learning, Canvass AI continuously trains the AI models, empowering operators to improve OEE by addressing industrial manufacturing’s most important AI use cases, such as anomaly detection, asset and process optimization, defect part detection, asset failure prediction, forecasting, and to model what-if scenarios.
FE: Does Canvass AI work with existing process control software, e.g., Rockwell, Ignition, etc.?
Malik: Yes. Canvass uses the edge gateway, which supports modern protocols like MQTT, to connect data from process control systems, like Rockwell, Ignition, etc. Canvass AI also offers integrations with other systems like OSIsoft PI. The good news is that there are many communication bridge options that allow manufacturers to connect a particular device to our gateway—whether it is using our REST API directly (allowing you to create your own custom integration) or a file-based data transfer over SFTP.
FE: What are the requirements (sensors, PLCs, software, etc.)?
Malik: The only hard and fast requirement is that a manufacturer has access to relevant process data they want to get insights from. Ideally, this is a historical data source, such as a process historian like OSIsoft PI. Once a manufacturer is ready to launch their AI/ML use case into their day-to-day operations, they will need to stream data to and from the platform.
FE: How does Canvass support MES and ERP systems? Is it meant to replace MES?
Malik: The Canvass AI Platform works alongside existing process control and business planning tools. Predictive insights can be used to make better data-driven decisions and complement rather than replace systems, such as MES. For example, manufacturers can use their ERP or MES as a source of raw data to generate those insights, in combination with data from control systems and historians.
FE: While Canvass handles production systems, is it meant to replace a PdM (predictive maintenance) system?
Malik: It depends. In some cases, Canvass may replace a PdM that doesn’t handle a use case’s complexity. In other cases, Canvass can complement a PdM’s rules-based model with predictive maintenance signals, which are based on a wider set of process data.
FE: Are you working with any food companies? What applications?
Malik: Some of the world’s largest food and beverage companies are using Canvass AI to improve the quality of their products, reduce energy costs and improve yield. Some use cases include:
Optimizing processes to increase yield: Using Canvass AI, a Fortune 100 food production company is improving its animal feed line’s profitability. The company utilizes machine-learning-based predictive analytics to optimize the fermentation process on a time basis to reduce batch cycle duration and improve batch consistency. As a result, the business unit has increased asset utilization, allowing the operations team to produce more animal additive feed without additional capital expenditure.
Improving energy efficiency and reducing carbon emissions: A North American food ingredient manufacturer introduced AI to optimize their plant’s energy production—with the aim to forecast the optimal power and steam generation required while minimizing gas usage. By optimizing multiple gas turbines, the company has lowered fuel costs and considerably reduced the plant’s greenhouse-gas emissions.
Optimizing processes to increase quality and reduce waste: Another food plant is using AI to improve the quality and shelf life of one of its grain products by ensuring consistent drying and optimizing the moisture level composition of the product. Previously, the operations team would need to perform manual checks on the moisture level every two hours. However, with Canvass AI, the operations team now predicts the moisture content across different time intervals and can adjust the various temperature and pressure set points required to accomplish the desired moisture content. As a result, the production line has reduced moisture level variance and improved the production line’s overall quality.
Improving quality and batch consistency of coffee bean roasting: A leading coffee bean roaster implemented AI to help them optimize the roasting time and temperature in order to achieve the required roasted bean color—which is a key indicator to ensure the bean matches the required aroma and taste. Prior to using Canvass AI, the coffee roaster would estimate the required settings, which would need multiple adjustments in water levels, roasting time length and temperature to get the color and moisture with the desired range. With Canvass AI, the operator can now predict the optimal settings, therefore, reducing wastage and rework and increasing their production throughput.
FE: What are your visions for the future?
Malik: I’m really excited about seeing real-world applications of AI for sustainability. Our customers are now using AI to drive their sustainability ambitions, innovating their operations to reduce waste and their carbon footprint. For a long time sustainability was seen as next year’s problem, but if one good thing came out of COVID-19, it showed how innovative manufacturing can be when necessity dictates. Now there is real momentum on how manufacturers can address their sustainability goals in a meaningful way. I see Canvass AI being a huge part of how they can accelerate this and create a sustainable future for us all.
For more information, visit Canvass Analytics.
