Big Data Analytics for Manufacturing - OptimalPlus

Cloudy with a chance of better analytics

A few weeks ago I discussed the steps needed to become a data-driven manufacturer. This was a high-level overview of the process of collecting, analyzing and acting upon data from the manufacturing floor.
Collecting data from sensors embedded within the manufacturing line can provide great insights into the efficiency of the manufacturing process itself – are we really utilizing each machine’s production threshold to its fullest? Can we optimize the yield by displacing one machine or adding more machines? Are there machines which are more prone to creating faulty outcomes? and so on.
This kind of data, when aggregated and analyzed, can of course be extremely valuable to any manufacturing company, but this manufacturing process is just a vehicle to creating a product, so we should remind ourselves that the more important – and even critical – questions should be asked in relation to the product, not the manufacturing process.

During the manufacturing process, the product goes through several machines, so a “per-machine” data gathering and analysis (which we, of course, provide) would most likely not give us the real insight on what went wrong during the process of manufacturing the product. There are sets of values which should be tracked, which are collected from the manufacturing machines themselves, but are put in context as part of a larger process of “product making”. Reliability, safety, quality, performance, consistency – these are all extremely important measurements of a product’s robustness and value, which would ultimately decide the fate of the product when in use.

Some of them, if not all of them, are directly derived from the process in which the product is manufactured, but in most cases the manufacturer is either a separate entity from the product vendor, or the manufacturing division is part of the product vendor organization, but it does not leverage its ability to potentially collect and share the relevant data for the “product” divisions (such as QA, design, customer support, logistics, etc.).

This is the difference between conducting “Machine Analytics” and producing “Product Analytics”, and while we at Optimal Plus are dealing with both, we passionately believe there’s a huge void in the market for Product Analytics – which we aim to fill.
In this diagram you can see a visualization of the data collection and analysis relevant to “Product Analytics” we offer today.

Diagram 1: System Architecture – Automotive & Electronics

The Manufacturing Execution System (MES) receives data from sensors attached to the various machines involved in the manufacturing process, and additional data from other equipment and data sources (internal or 3rd party oriented). The data is shipped through Optimal Plus’ agents; aggregated at Optimal Plus’ Edge repository; aggregated data from the various locations is moved to and analyzed at Optimal Plus’ Central repository; the analytics’ insights are communicated to and enhanced by the customer’s PLM, ERP, CRM and other management platforms; and everything is accessible through the Optimal Plus dashboard, Portal+.

Leveraging clouds to reach for the sky

As in many other domains, vendors in the semiconductors and electronics domains are realizing the benefits of moving more of their analytics infrastructure to the cloud, and take advantage of its elasticity; enablement of constantly updated 3rd party infrastructure, applications and specific modules which can plug-in to their own processes; cybersecurity expertise, etc.

This is where the “Hyperscalers” – cloud infrastructure vendors such as AWS, Microsoft’s Azure, Google Cloud Platform and others – are getting into the Industrial IoT (IIoT) or “Industry 4.0” game. In addition to providing their traditional claim to fame – being a more cost-effective way of storing and moving data within a cloud environment – these hyperscalers began their foray into the manufacturing analytics world, and using their own developed tools and through partnership with other domain-specific vendors, they provide modules which perform or support analytics processes done by domain expert vendors.

But hyperscalers know they can’t be the best of everything, so they want to be the best in something. If they can work with the best in class when it comes to the application layer, they can position themselves as the best in class when it comes to the incremental offering of the infrastructure + industry-specific applications.

Hyperscaling our cloud operations with a hyperscaler

What does it all mean? Well, while we pride ourselves in being great at what we do, we realize that we can’t be the best in everything, so we partner with vendors which we believe have mastered the needed infrastructure and processes, and combine our subject-matter-expertise and use-case customized products, with their industry-leading prevalent and trusted platforms.
Our passion for Product Analytics, combined with the fact that there are great vendors of analytics-enhancing infrastructure vendors – such as the aforementioned Hyperscalers – cumulates into this current step in our ongoing journey to enable the manufacturing sector to improve both production yield and product quality.

We recently had a chance to test this vision with our friends at AWS, as we ran a POC in the world of manufacturing insights for the electronics sector. In this POC we created an end-to-end process of collecting data from IoT sensors on-premise, preparing the data for analysis, shipping the data onto the cloud, conducting various process on that data, and creating actionable insights and alerts – where most of the cloud-related processes were done with AWS Iot-specific modules.
Did it succeed? Well, I wouldn’t be telling you about it if it didn’t , right? 

Sign up for updates, news, use cases, best practices and events

OptimalPlus is the global leader in lifecycle analytics solutions for the automotive, semiconductors and electronics industries, serving tier-1 suppliers and OEMs.