Covering Disruptive Technology Powering Business in The Digital Age

Authored by: Yee Bing Hong, Business Development Manager ­­− Keysight Technologies

In the electronics manufacturing industry, Original Equipment Manufacturers (OEM) and Contract Manufacturers (CM) are always trying to optimise their manufacturing processes to improve efficiency and quality and reduce wastage. With the introduction of Industry 4.0 enablers, the digital transformation to the smart factory presents many possible areas of opportunity to invest in. It would be obvious to consider transforming the digital backbone of a typical factory.

For decades, the digital backbones of factories have been Manufacturing Execution System (MES) solutions. MES has been invaluable for monitoring the process status of products being manufactured as they move through the line, providing visibility, tracking, tracing and high-level control. The summary data that is acquired from these processes have been designed to provide just what is needed yet make the payloads light and efficient,  consisting, for example, of serial numbers, stage in production, time stamps, process pass or fails, etc.

Database technology was kept simple and straightforward, typically using traditional Relational Database Management System (RDMS) data warehouses, rather than following the current trend of complex big data no-SQL systems. All these are a consequence of addressing simpler use cases, where the most challenging are at most sampled complex statistical computation for quality control best practices and standards.

In the electronic boards manufacturing sector specifically, quality gates such as In-Circuit Test (ICT) and Functional Testing (FT) systems are ubiquitously used to ensure only good parts get shipped out. However, these test systems had very high-level summary data that were visible in the MES systems. Some of the summarised data would include metrics such as First Pass Yield (FPY), Final Yield (FY) Statistical Process Control (SPC), downtime, test time, test volume and so on.

Whenever there is a low FPY incident or if there was a need to improve existing FPY of a product, an in-depth offline root cause analysis is usually undertaken by engineers. The deep analysis may point the root cause at some process upstream, which was well-caught at test. A common practice will be to extract data logs manually from the test systems. Then the data will need to be cleaned and transformed into something that can be used in an offline tool such as Excel or other similar tools.

This offline analysis solution is a slow and painful method. In addition, repeating the extraction, cleansing and transformation of the test data logs is time-consuming and prone to human error. To make things worse, importing large amounts of test data into any tool in one attempt usually takes a long time. The feeling of waiting hours for the import process to finish and then seeing an error prompt is all too familiar for many.

With this concern from the existing solutions in most factories, there is a need to close the gap for current analytical solutions, specifically for In-Circuit-Testers and Functional Test (FT) systems if the manufacturers intend digital transformation for the improvement of yields and quality.

In order to close the gap, manufacturers need to consider big data analytics solutions that are scalable, easy to deploy and fast. PathWave Manufacturing Analytics (PMA) is Keysight’s industry-leading Industry 4.0 analytical solution that enables electronics manufacturers to upgrade their factory to the next smart digital transformed factory by providing real-time advanced data analytics on the massive raw test data logs. It handles source-to-target with Keysight’s innovative big data stack that uses the best big data components such as KX’s kdb+. What makes it truly differentiated is that PMA combines years of test and measurement knowledge in manufacturing test and emerging data science to cater to electronic manufacturing use cases, especially in ICT and FT systems.

By adopting a big data analytical platform, actionable insights can be generated from seemingly wasted massive amount of raw test data that are being produced every second. Some of these advanced analytical insights include probe degradation predictions, false failures trends and anomaly detection, to name a few.

After establishing the need to adopt a big data advanced analytics platform in a smart factory, the first question is whether to build an internal solution or explore matching solutions from solution providers. It is quite often the case where there in the DIY mindset of engineers, that they want to try out and analyse the machine data themselves. But this is easier said than done, and is it worth developing your own manufacturing analytics platform? Will you get a good return on investment (ROI)?

Here are some issues you may need to address, to make the right decision.

Building Your Own: Time, Money, Risk

Building your own new big data advanced analytics platform requires a sizeable team of experienced specialists, requiring a mix of computer science, data science, full-stack developers, UX designers, big data engineers, testers and product owners to work closely with test engineers that have the domain knowledge of the use cases. Aside from the time spent on design and development, the time spent on hiring or upskilling as to be taken into account as well.

“Manufacturers need to consider big data analytics solutions that are scalable, easy to deploy and fast. PathWave Manufacturing Analytics (PMA) is Keysight’s industry-leading Industry 4.0 analytical solution that enables electronics manufacturers to upgrade their factory to the next smart digital transformed factory.”

After developing and deploying the first iteration of the solution, which may take anywhere from 6 to 12 months, the development teams will need to be reassigned to other projects or even let go if under contract. However, a smaller team will still need to be retained for continuous maintenance, enhancement, improvement and support. Whenever there is a need for major enhancement, there will be the challenge of finding the resources to support the work.

On the other hand, choosing a suitable solution from a solution provider, will involve probably a week or two of setting up and maybe a few hours of training, and all your KPIs and analyses will be yours to work with. Valuable time can be better utilised to focus on core business operations and improve organisational efficiency.

Specialised solution providers such as Keysight can provide the after-sales service and support through onboarding, customer success consultancy, new feature enhancements, bug fixes and high availability.

Connecting the Different Types of Electronics Test Systems on the Production Floor

It is common to have many flavours of test system in the factory. Every test system usually has its own format of data outputs and test results and is dependent on the individual engineer developing the test plan or script. The granularity of the test data can range from common pass/fail data, to test data with a specific Bin/Error code.

More often than not, the test data logs are missing some types of data that would be useful for big data advanced analytics! This will potentially create a permanent gap between the data available and the desired outcome and insights looked for, unless changes are made to the original test plans. Of course, that is if the original test engineer is still around or at least has the test plan well-documented before the initial final release.

“Building your own new big data advanced analytics platform requires a sizeable team of experienced specialists, requiring a mix of computer science, data science, full-stack developers, UX designers, big data engineers, testers and product owners to work closely with test engineers that have the domain knowledge of the use cases.”

This is why it is very important to understand the raw ‘datascape’ of test systems out on the floor, and whether there are any potential gaps between what is available and what your desired analytical outcomes and insights are. We have seen many cases where the attitude of ‘bring whatever we have into the data lake first, and then we discover what we can do’ has resulted in massive failures and money sinks. Rather, ‘begin with the end in mind’ and you will probably get your actionable insights.

Bigger Data Means Bigger Truths

How do you ensure that you are on par with your competitors, in production efficiency and quality?

It is not possible to answer that question unless without getting into a competitors’ firewall.

But off-the-shelf manufacturing and analytical solutions can give you a sense of what the industry is doing and what insights are being generated for, for instance, anomaly alert scoring and test coverage monitoring amongst others. By understanding, developing and deploying industrywide use cases with a lot of industry inputs and validation, you can be assured you will be getting the best insights that any analytic solution can offer. Bigger data means bigger truths.

Moreover, by obtaining continuous support and constantly conducting two-way engagements, you can see new use cases and industrial trends, in order to continuously improve product in a truly agile fashion.

(0)(0)