Computer Vision Advances Zero-Defect Manufacturing

A bald man in a suit smiles for the camera.

Cloud Computing Evolves: An Interview with Mats Johansson

By G C Network | June 6, 2019

Recently, Ericsson Digital released an amazing report on Edge Computing and 5G. In it, they explained how distributed cloud computing is paving the way for the future of network communications. They…

Rexroth unveils rexroth rexroth rexroth rexroth rex.

The IoT Nexus: Bosch Connected World 2019 in Berlin

By G C Network | May 11, 2019

Next week, I will be influencing #LikeABosch as I accept an invitation from the company to attend Bosch ConnectedWorld 2019 (BCW19) in Berlin, Germany. This is one of the world’s largest international…

A group of people sitting at a conference table.

Survive and Thrive With Digital Transformation

By G C Network | April 17, 2019

First cloud computing then multi-cloud. How can we get ahead of this digital transformation nightmare? These are the laments heard in conference rooms and board meeting around the world. While…

A cartoon man standing next to a white tesla model 3.

The “George Jetson” of Today

By G C Network | April 13, 2019

  He grew up in Silicon Valley, landed his first job at Apple Computers, was introduced to Nobel Prize winners by his dad and today, he takes a self-driving car…

A city at night with the words 5g on it.

MWC19: Where Telecommunications and Cloud Meet

By G C Network | March 23, 2019

As a cloud solution architect, my passion is learning the details about how cloud computing uniquely supports specific business cases. This curiosity is what drove my excitement when Ericsson invited…

Tulane university school of professional advancement logo.

Tulane University SoPA Selects “Architechting Cloud Computing Solutions”

By G C Network | February 16, 2019

Last week, Packt Publishing announced that “Architecting Cloud Computing Solutions” by Kevin L. Jackson and Scott Goessling was selected for use by the Tulane University School of Professional Advancement, Applied Computing Systems & Technology Program as the textbook for…

A group of doctors looking at a computer screen.

5G Wireless Technology Connecting Healthcare

By G C Network | February 16, 2019

Healthcare is in the middle of massive change. Called digital transformation by many, this term describes the industry’s pursuit of the many promises offered by connected patients, connected caregivers, and…

A woman wearing a red jacket and necklace.

Maria Lensing: The Network Platform for Healthcare’s Future

By G C Network | February 16, 2019

As a girl, Maria and her family traveled to Memphis, Tennessee to get cancer treatment for her sick brother. The miracle she observed, as the healthcare providers saved her brother’s…

An image of a network of dots and lines.

How “Big Iron” Does “Big Regulation”

By G C Network | January 10, 2019

According to Verizon, there were over there were over 53,000 security incidents in 2017, with over 2,200 of those identified as confirm data breaches. A Ponemon Institute study also showed…

Two different signs that are side by side.

Mainframe Synergies for Digital Transformation

By G C Network | January 10, 2019

In July  of 2018, Broadcom announced its intentions to acquire CA Technologies. In the press release, Hock Tan, President and Chief Executive Officer of Broadcom, said: “This transaction represents an…

by Kevin L. Jackson

Electronics manufacturers operate in a challenging environment. It’s hard enough to keep up with the ever-accelerating rate of change in the industry. Now customers want increasingly specialized product variations in less time and of higher quality.

Meeting this demand for increased product variation can seriously impact the bottom line. Such variability increases warranty cost that, in the US, currently averages 2.7 percent of electronics manufacturers’ revenue. This means that for every $1 billion in revenue, a company spends $27 million to support its product warranties.

Case Study: HPE Server Manufacturing Challenge

These factors explain why Hewlett Packard Enterprise (HPE), a leading manufacturer of servers and other IT hardware, is undertaking an Industry 4.0 strategy to modernize and automate assembly processes.

A key part of the process it wanted to automate is server assembly quality assurance, which was being done manually by quality operators. This labor-intensive process is prone to error due to human eye fatigue and the inability of quality operators to catch critical defects.

The company’s portfolio of server products is complex and highly customized. These characteristics made it difficult to find a machine vision system that could handle the product mix.

This situation is hardly unusual. According to Kemal Levi, Founder and CEO of Relimetrics, there is “a strong demand for computer vision to replace manual visual inspections. Yet, due to a high production variability, particularly in the case of discrete manufacturing, computer vision systems today are not able to keep up with the rate of change in configurations.â€

Quality Audit by Relimetrics

To address these concerns, HPE is implementing Relimetrics Machine Vision Systems to automate end-of-line quality inspections at their server assembly lines. The Relimetrics Quality Audit Solution for Electronics (RELI-QA) is a proven Zero Defect Manufacturing (ZDM) solution that uses machine learning to achieve a high level of flexibility and scalability for computer vision systems (see Figure 1). The solution can be rapidly deployed, trained models from one line and data can be shared across lines, and existing camera hardware can be leveraged.

Figure 1. Relimetrics Quality Audit Solution for Electronics (RELI-QA) uses computer vision and machine learning to improve quality.

With this innovative approach to ZDM, the following can be achieved:

  • 40-second inspection time for 80 components in an entire server
  • Greater than 99.9 percent probability of defect detection
  • 50 percent cost reduction in the company’s quality audit costs
  • 80 percent cost reduction in the company’s final testing and rework

By implementing this solution, an IT electronics manufacturer producing 30,000 servers per month can reduce the cost of rework by half a million euros per manufacturing line per year. What’s more, the solution slashes the time to pinpoint the root cause of quality issues from days to minutes.

The Value of ZDM

HPE is not unique in its move toward automating quality audit (QA) as part of a larger ZDM strategy (see Figure 2).

Figure 2. The Zero-Defect Manufacturing concept that supports Industry 4.0.

Many other manufacturers are pursuing this approach thanks to valuable aspects of ZDM such as:

  • Its ability to considerably reduce the costs of handling defective products
  • Removing all redundant elements within a manufacturing process
  • Reducing scrap production and associated expenditures
  • Continuous production supply chain improvement
  • Increased safety and customer satisfaction

A recent International Journal of Production Research study stated that the evolution of Industry 4.0 has made ZDM implementation easier due to the availability of large data sets required for modern machine learning techniques. Increases in computer power and storage have combined with decreasing cost of sensors to the point that it now makes ZDM a much more cost-effective option.

A Practical ZDM Solution

The Relimetrics platform demonstrates the power of these advances. The solution can quickly evolve and adapt to changes in production parts and configurations, and enable the customer to sustain a high (>99.9 percent) level of accuracy by filtering inspection results automatically and continuously updating deep-learning models at the edge using advanced HPE edge-computing technology.

The advantage of this approach is the reduction of downtime associated with reconfiguring algorithms in high production variability manufacturing environments without any expert knowledge required. Levi explained, “Existing machine vision solutions cannot handle high production variability. Either manual intervention is required to reconfigure the algorithms or dependency on the tech vendor for updates leading to downtimes.â€

In fact, in today’s QA, the downtime associated with reconfiguring algorithms is a recurring problem experienced by many plants owned by a global manufacturer. Often, the same products or parts are being manufactured on different lines in different geographies. But these plants are not able to actively share the QA learnings with one another and take advantage of one another’s data.

Another advantage of the Relimetrics approach is its ability to leverage data produced at different lines and plants to deliver a comprehensive QA solution for a customer’s global manufacturing environment. Levi also explained, “Today, the learnings from configuration changes in, let’s say, a plant in Singapore cannot be immediately transferred to another plant in Mexico. Whereas with Relimetrics solution, the learnings from one plant can be immediately shared with another plant in another part of the world.â€

HPE Servers Inspected by HPE Servers

These ZDM accomplishments are made possible by using advanced Intel® edge-computing technologies. Running computationally intensive machine-learning calculations at the edge enables error detection accuracy of more than 99.9 percent.

In an amusing twist, the Relimetrics QA system used to inspect the server production line is itself based on HPE servers. Specifically, Relimetrics uses HPE Edgeline EL4000 servers based on powerful Intel® Xeon® processors.

But the critical contributions from Intel go beyond hardware. Levi said, “Intel provides the software and the architecture that enables Relimetrics to make the best use of their CPUs. Their technology is critical to the quality assurance solution because it enables us to run computationally intensive calculations at high speed.â€

Originally published by on May 29, 2020.

Sponsored by Intel Corporation

Posted in