This deep tech startup uses digital twins to predict performance on shopfloor

Artificial intelligence has found a wide appetite for adoption and innovation in the consumer and enterprise ecosystem. But is nascent in manufacturing, which is surprising as it frontier or deep tech has the potential to solve the unique problems of the sector. Over the course of last year, manufacturing sector’s need to remain in near perfect condition without a hint of plant level bottlenecks is apparent. Enter SwitchOn, with its solution of creating digital twins to predict performance and enable real time intelligence from shop to top floor.

SwitchOn was founded in 2017 by brothers-Aniruddha and Avra Banerjee. At this point, Aniruddha was already working in the US with a multinational software solution provider, while Avra was with an aerospace startup back home. After the former was confident of having assimilated enough technology and business development knowledge and latter’s expertise in product development and AI, led to the conception and development of SwitchOn.

A curious case of quality inspection:

Automation has already upended manufacturing industry, strangely quality inspection is still a manual process in places. Because of the high automation speed) almost 10-12 products/second) there is a possibility of product quality defect. Also, manual inspection is time consuming and error prone. Therefore, a product with defect is found later in the chain or at times shipped.

The collateral damage of this occurrence is a bad customer experience. But the problem is magnified with the automotive industry, one SwitchOn has customers in, says Banerjee. A manufacturing defect could lead to accident and/or financial damage. (Remember when Royal Enfield issued recall of its Himalayan, Interceptor 650 and Continental GT 650 bikes over brake components corrosion that could have potentially led to brake failure.)

This makes a good case for automation of inspection on the shop floor—a prime forte of SwitchOn. “We capture high frequency data from the likes of stamping and welding press. Use a combination of AI and patented platform Xavier^TM that runs these algorithms and figure out the defects at the speed of 10-12 products/second,” says Banerjee. It’s completely automated in terms of the rejection system as well. “We can reject the products in the line at that speed.”

He also lists the benefits that include but not restricted to the guaranteeing the quality of every single part. “It helps drill down the customer complaints and helps create unique differentiator with respect to quality control. It is augmenting the kind of operators and plant personnel by also giving them the ability of visualizing the whole plant in the form of a digital twin,” he adds.

The winning technology: Digital Twin

The underlying concept for SwitchOn’s unique solution is digital twin. In Banerjee’s words, “We give them a digital twin of the entire plant and tell them exactly where the problem is.”

SwitchOn defines the digital twin as a combination of the asset specific information listing out the ideal operating behaviour, coupled with the high frequency data. A bunch of operating modes of the asset is figured using a patented technology. This can determine what the asset is making at a particular point. The data is configurable depending on the asset being worked with and rest of the pipeline is same, says Banerjee.

The next step is figuring out asset failure or problems of quality in the manufacturing. “There are two kinds of control loop: Hardware level integration work with rejection system, done through PLC (A module triggers rejection function as soon as an error is found) and second, triggers the rejection mechanism. second, trigger 4m workflow. This reduces failed the amount of products and turnaround time,” says Banerjee.

Edge computing: A competitive differentiator

Banerjee throws light on the difficulty associated with procuring high frequency data. “We take sample data of about 2000 samples a second. It means we have humongous amount of data that goes into the platform. It means we need to come up with our own hardware to do that kind of processing on the fly,”

The edge computing has made it possible to crunch high compute into a wallet size device. “Our platform, Xavier, makes end-to-end process a matter of weeks rather than months,” he adds.

Additionally, knowing about assets is as important as gathering big data, believes Banerjee. Especially when you find solutions with deep tech it is essential you have subject matter expertise. “We choose to verticalize the platform to only a few industries and use cases. That led us to understand these machines much more. The critical missing piece is asset specific expertise to make any of that happen,” he adds.

Building customer base:

SwitchOn is a graduate of Axilor Accelerator Program and has been mentored by NASSCOM Deep Tech Club. It has raised a seed funding of $1 million from Pi Ventures, Axilor Ventures and The Chennai Angels. Banerjee says SwitchOn’s priority is to build deep enough customer base in India, while also expanding the small base of customers in the US.

Banerjee believes that the Indian consumer is becoming quality conscious with the generation of millennials coming in with a world class outlook and as the nation is also stepping into a lot of export in terms of product. “That is driving big tailwind onto the quality inspection and automation design trends. The idea is to always build from India for global audience and that will be a differentiator in terms of global landscape,” he concludes.