Automating 20/20 Vision

In 2013 a company called Doorbot started selling computer vision for home security. There were already hundreds of home security systems on the market but what was different about this one is they leveraged something called computer vision. Where other companies would just install motion detectors for low cost systems or video surveillance cameras at the very high end, Doorbot combined both. Because computer vision was getting smarter and cheaper, their customers, DIY homeowners, could set the boundaries of where the Doorbot system could look and activate when motion was detected. Left, right, near or far it could be trained to know where to look and when to start recording. And instead of human looking at a monitor, the system pushed a notification to the homeowner when there’s something to look at. The company, shortly after appearing on Shark Tank, rebranded as Ring and the rest is history. 

Much like Doorbot, Atollogy founder Rob Schoenthaler saw a crowded field of technology offerings in the manufacturing space and the low cost and simplicity of computer vision as an opportunity to find a need and fill it. He went to work on a solution that would address the biggest questions manufacturers struggle with today: “what’s my cost to make this part?” “am I utilizing my equipment as best I can?” “Do I have the right amount of labor deployed?” 

Traditionally, companies employ an Industrial Engineer to watch and learn their operations over a course of months. The Engineer records all the activities relating to the movement of people from machine to machine. How many are at each workstation and for how long. He records what they are doing such as actively working, waiting for material or waiting for maintenance. He records the status of machines by either knowing what the Andon lights are telling him or by learning over time when its running or not and why. More advanced industrial engineering engagements also target productivity, measuring how many parts are produced by machine, person, team, or shift and then break this down by time and or associated part number.

At the end of the engagement the engineer and his team have reams of data that they pour over to find causality and trends to answer questions such as: what’s the overall OEE (Overall Equipment Effectiveness), what’s the best mix of labor rate by shift or machine, how long is each machine down and how long does it take to get it back up and why. What are my top reasons for not producing a part by machine, shift, day or SKU? The output of the Industrial Engineers work is a set of recommendations for the company to make process improvements in the three key areas of machines, people, and inventory. The engagements can take months or even quarters, cost into the 6 figures and are a one-time fix to a very fluid manufacturing process. 

Now imagine everything above could be done without a human – and this is the business capability Atollogy developed using computer vision. With the help of several world class manufacturing companies, they developed the Digital Industrial Engineer (DigIE). An Industrial Engineer that works 24 x 7 x 365, observes people, machines, and materials but instead of compiling reams of data over a long period of time, DigIE learns patterns and processes within weeks and converts the physical world it is observing into a digital one in real time. It’s hard to manage or improve a process that has not been measured. DigIE makes sure users can see their digital operations in weeks in order to get a baseline on their operations, and over the course of a month are able to find trends and causality through built in analytics that would mimic the findings of the process engineer. 

The best part of DigIE is it was purpose built for DIY use. It requires no integration to systems or even a network connection. It’s delivered as a fully autonomous and movable kit that can be set-up and moved by the customer. DigIE is trained to look at your areas of interest for real time reporting. Like Ring, the DigIE never takes a break, it never blinks and the algorithms are not subjective.

Author:

 

Dan Murphy

Atollogy Eastern Region Lead

dan@atollogy.com