Intel has begun deploying internet of things sensors to improve the efficiency of its own chip manufacturing plants, encouraging staff to innovate with its low-cost Galileo computing boards.
The semiconductor firm relies on huge fabrication plants across the world to manufacture its range of processors, with complex production methods involving the use of billions of dollars worth of equipment.
This work generates around a million data points as Intel monitors the status of the facility, tools, chemicals and gases, as well as the quality of production wafers and finished products. However the firm is seeking to increase the amount of data it collates through more sensors to find ways improve its operations.
“There is huge data already in our factories, but this is tip of the iceberg for us. We are looking at going from a million to several million data points just by letting people to get their hands dirty and solve the little problems that bother them every day," said Intel business analyst Dan McCulley.
“So [we are going from] big data in our factories to even bigger data because these internet of things connected systems allow us to quickly take that equipment and bring it to the web.”
The 'maker' scheme was launched in October last year, with Galileo devices provided to staff across its global operations. Based on Intel’s Quark and Pentium system on chip (SoC) processors, the Arduino-compatible Galileo computing board is aimed at students, enthusiasts and ‘maker’ communities, allowing ideas to be quickly tested.
Since the start of the programme, 150 prototypes have been launched, McCulley said.
Solving manufacturing problems
The computing boards are used in a variety of different ways. One is to attach temperature sensors to robots which carry wafers around the fab, allowing staff to track temperature and air pressure. This allows Intel to use multiple mobile thermostats that continually collect data, rather than just a small number of static sensors.
Another project is to install sensors in tool boxes, helping to prevent equipment being lost around the manufacturing plants - costing around $35,000 a year at each fab. This is done by requiring staff to provide their identity when accessing tools, with light sensors in the box detecting that equipment has been taken, and logging this information to individuals, providing traceability.
McClulley said that using Galileo for mobile water quality testing would cost less than $5,000 per system, rather than spending on automated systems that require spending between $50K to $100K.
The company has also begun to use Galileo for sensing in testing environments.
“When we do burn-in for our chips, we do it at high temperature and humidity. We typically have one thermostat in the test facility. Using a Galileo, now we can put 15 or 20 thermostats into the test chamber,” said McCulley.
“This goes back to the big data concept - instead of having one data point to correlate to the quality of the chips, I can now correlate every single rack to its individual characteristics and performance. You get better data, and better results."
Moving from proof of concept
The 150 projects are currently at prototype stage, but Intel is planning to roll out the most successful technologies more widely across its business.
McCulley said that for large scale use of IoT sensors across is business, it is likely that Intel would switch from using the Galileo boards to an industrial gateway controller - essentially offering more reliability with extra compute power and redundancy - but would only require a day to convert the systems.
It is could also replace Galileo boards with Intel's smaller Edison device for certain use cases when moving to mass production.
One of the main benefits of using the development computer board is the ability to address organisational inertia by exhibiting the improvements that are possible for a low cost, McCulley said.
“What we have found is that you only have to do a couple of these and management gets it. The Galileo allows me to quickly prototype and solve a problem before.”
“You can show it works - it is so much easier to have that conversation with someone about spending $300 when I have already shown them I can save $35,000.”