With the Industrial Internet of Things (IIoT) and via process-oriented analytics, smart manufacturing environments are positioned to make enormous gains in reducing operating costs, better uptime and improved asset performance management.
But what about the quality of the products that originate and transverse throughout this new IIoT infrastructure? In today’s fast-paced media environment, a product’s quality shortcomings can quickly become headline news, threatening financial implications for companies of all sizes, ranging from recall costs, to brand damage and to future product sales.
What value does product-oriented analytics bring in today’s semiconductor and electronics manufacturing industries? Being part of a big data analytics software company I would like to share some insights with you.
Product quality in the IIoT era: Listen to the voice of your product!
In the quest for IIoT process-based optimization, companies should also consider complementary solutions that can ensure individual product quality. Whether the final product is destined for an automotive ADAS system, a wearable heart monitor or a ubiquitous smartphone, product quality vigilance must be paramount. In the best of manufacturing environments, products may still have undetected quality issues . To harness the full value of the IIoT for OEMs, manufacturers and end users, it is becoming increasingly important to perform product analytics in addition to process analytics.
The application of product analytics enables companies to hear the “voice of their product”. It makes it possible to significantly improve product quality and at the same time enhance existing process analytics to further improve operational efficiency and product yield. Deep product analytics within the IIoT-based smart factory represents the next level in the promise of the IIoT. Let’s take a closer look at how product analytics adds unique value. What does product analytics provide that process analytics doesn’t?
Product analytics vs. process analytics
The most common element to all electronic systems is the semiconductor chips that drive the functionality of the device. Analytics directed at optimizing machine performance in the manufacturing and testing of these chips does an exceptional job at ensuring peak manufacturing efficiency. However, process analytics only provides a very coarse-grained and limited assessment of product quality: good or bad.
Product analytics on the other hand, when performed on harmonized process and test data, can use the power of big data analytics to identify many nuances in manufacturing operations that are impossible to see in a binary test environment (good or bad). It is through these deep product analytics that quality, yield and productivity can be significantly improved.
I picked out some examples of product analytics insights for you that directly impact RMA reduction, yield improvement and brand protection:
Is a device tested as “good” really good?
One case that we see at almost every customer are devices that are shipped into the supply chain without being properly tested due to a tester “freeze”. This is a situation where multiple consecutive devices return the same parametric test result. These subsequent devices were not actually tested. But the test program recorded a result that was effectively “copied” from the prior device because of the freeze. Without a big data solution in place that can search through thousands of test results in real-time, it is next to impossible to find these test errors before the devices are shipped into the end market, resulting in a potential RMA situation.
Is a device tested as “bad” really bad?
Similar issues can happen with good devices being labeled as “bad”. A tester basically tests a device and determines if it is good or bad and then moves on to the next device. However, these machines don’t remember what the results were for a product line it was testing even a few short hours ago. In other words, a tester could be yielding devices at 95% at 10am, but then drop to a yield of 85% by 2pm. In most cases, it isn’t because incoming product quality dropped. These types of yield drops are often due to some sort of problem with the tester hardware itself. By analyzing yield data, product analytics can identify statistical variations, such as reduced yield, that can indicate when a tester may need earlier than expected maintenance. Identifying these types of gradual degradations in test results as soon as they begin to occur can help significantly reduce operational issues such as yield loss due to equipment problems. That cannot easily be predicted by process analytics alone.
Quality of chip compatibility or identifying grades of “good” and “bad”
Device “A” may function well when paired with Device “B” from Vendor “X”. And it may function poorly when paired with Device “B” from Vendor “Y” (a 2nd source manufacturer). In both cases, Device “B” successfully passed all the quality tests from their respective vendors. With the high-performance bar that every electronics manufacturer is trying to achieve with their products, big data product analytics can help companies distinguish between different grades of “good” devices. It helps to understand how they can affect overall system performance. A simple example of this is the use of slow DRAM memory with a fast processor in a PC. It is a waste of the fast processor to be paired with slow memory. Better to pair higher performing memory with higher performing processors and vice-versa. Many companies are now implementing manufacturing methodologies that include the “smart pairing” of devices in multi-chip modules to maximize system performance.
Establishing product “DNA”, traceability, and supply chain visibility
Another benefit of product analytics is to improve the overall efficiency of the supply chain, by providing full traceability across the semiconductor-electronics ecosystem, including the capture of product “genealogy” data from every semiconductor IC used in a design to the printed circuit board the devices are mounted on. The ability to visualize and act upon all your manufacturing data, regardless of the source is an essential benefit that the IIoT can provide.
The supply chain of a system with electronics content is very complex. It can generate a large amount of useful, but often-incompatible data. For example, wafers that are manufactured in foundries and sent to packaging houses. The packaged chips are then installed on boards, boards go into sub-systems, and these go into end-systems such as cars or phones. Each step is completed by a different company and sometimes by more than one company, and at disparate global locations. A multitude of tests are conducted throughout the manufacturing process using a variety of test equipment and producing massive amounts of test data that is unique to every company in the supply chain.
Fortunately, product analytics can act as a unifier in this complex supply chain. Test results collected throughout a global supply chain can be used to establish the “product DNA” for every individual device that is manufactured. The visibility provided by this DNA can help correlate between customer returns, the PC boards in the system, the semiconductor chips that reside on those boards, the company in the supply chain that manufactured any part of the final system, down to a specific tester that was used and the time the tests were run. These supply chain insights give companies unprecedented transparency and traceability into product genealogy at every level of the supply chain.
Alas, product analytics is hard to do
Process analytics is often focused on a specific manufacturing environment such as the machines on a factory floor, or the factory floors of a specific manufacturer. Contrary to this, good product analytics is all-encompassing in several dimensions – throughout the supply chain from chips to boards to systems, throughout the product lifecycle from design to new product introduction (NPI) to high-volume manufacturing. As such, product analytics can provide deeper levels of insight and much more value to a semiconductor or electronics brand owner, but is also much harder to do. Consider the following:
Sheer amount of data. Collecting, cleaning, managing, and analyzing data from the entire supply chain and during the entire product lifecycle can easily reach into the hundreds of terabytes, requiring true big data capabilities.
Single point of truth. Holistic product analytics require collecting and analyzing data from multiple suppliers, and from a variety of organizations (e.g. R&D, Test, Operations). Securely connecting multiple global companies requires the ability to harmonize all that data. It requires to create a “single point of truth” for all product data for the brand owner.
Process analytics is an important aspect of the semiconductor and electronics manufacturing industry. But to get the full potential value of the IIoT for these segments, it must be accompanied by product analytics. The ability to understand product test DNA and to be able to listen to the “voice of the product” throughout its lifecycle will enable semiconductor and electronics manufacturers to significantly improve the quality, yield and performance of the products that bear their name, and ultimately deliver greater protection for their brand.
Are you interested in learning more about using big data for product analytics? Get to know a semiconductor company standpoint and watch this video.
The post How to enhance product quality in the Industrial IoT era appeared first on Bosch ConnectedWorld Blog.