Hacker News Reader

Comment by contingencies | Hacker News Reader

Recently learned a lot about software, theory, process and management in the design, iteration and manufacturing of custom printed circuit boards.

Having climbed that mountain, I'm currently diving in to industrial automation: not software automation, but both off the shelf and custom robotics and factory equipment providing manufacturing 'unit operations' for working with metals and polymers. It's quite impressive how little integration currently exists between products from different vendors. There is a huge opportunity here for a software-world startup. Competing concerns include safety, reliability, problem detection, automated issue resolution, machine servicing, tool longevity, vendor desire for lock-in, supplier desire for lock-in, noise, dust, finish, speed, logistics, HR, scalability, cost, etc.

View on HN

I delved into this exact area in the last startup I was at. I worked pretty extensively with Kuka 6 axis industrial robots and spent a lot of time programming them as well as integrating them with our custom hardware. Totally agree that there is a big opportunity for a unified software layer on top of all this hardware.

The system I ended up building allowed users to monitor the robot from the cloud, including telemetry such as end effector sensors and axis positions, as well as uptime. It also allowed steaming tool paths from the web platform to the robots directly with appropriate safety measures such as manual intervention required to start it. Would love to generalise this sort of thing and create a turnkey solution that allowed people to focus on solving their actual robotics problems, not spend all their time glueing together disparate components.

Sounds interesting, although single vendor is much easier than multi-vendor, especially when you throw in parallel purchasing decisions on short timelines and both execute and receive sales and tech support in Chinese...

A generalized solution necessarily has a strong spatial component. Most vendor motion planning applications seem to be either custom and cell/unit-operation specific or are expensive specialist general applications with high complexity that often exist as plugins to CAM tools like SolidWorks.

I don't have the time and resources to acquire and delve in to those, so my approach has been to accept the fact and go ghetto with cheap local hardware that comes without motion planning or integration support, observe existing industry processes, and attempt to self-build iteratively. Luckily I've already built a team to produce custom robotics cheaply and quickly and recently added an experienced production manager from the automotive sector, so we have an advantage in this space.

Any suggested reading/resource list ?

Not really, someone could perhaps write a modern book approaching manufacturing from a software background. My feel-it-out approach has been a multi-year gentle introduction combining reading and MOOCs across logistics, trade-specific fabrication manuals, metrology and an understanding of fundamental mathematical approaches to the space and evaluating their existing implementations: chiefly multivariate linear optimization and Google's OR Tools. I also maintain a custom supply chain automation platform for our business and now have multiple years of experience working with a range of third party fabricators (many orders per week) across a few processes and material types which has given me relatively strong insight in to a broad range of challenges or issues that can emerge in fabrication-related process interfaces: bad specs, tolerances, shipping damage, alignment, test overheads, supplier reliability fluctuations, shipping issues (carrier bankruptcies!), missing packages, importance of receivable/inventories, reasoning behind anomalous unit operation timings, importance of maintenance, fallback, training and safety, insurance costs, true costs of manual programming, benefits of insourcing, lesser-known unit operations can save a ton of money...

Reply on news.ycombinator.com