Like other industrial manufacturers, the IBM Systems Manufacturing division looks to use AI-powered automation to boost efficiency while maintaining quality. The Systems team decided to use IBM's own Industry 4.0 solutions to automate and accelerate quality-inspection processes.


The Systems Manufacturing team developed an AI-powered, automated inspection solution that runs at the edge for maximum efficiency. The solution combines two of IBM’s leading AI and edge solutions, IBM Maximo® Visual Inspection and IBM Edge Application Manager software.


5x increase

in inspection efficiency

20% reduction

in false-positive defect detections

20% cost savings

on software maintenance

Realizing the promise of Industry 4.0

Manufacturers today recognize the promise of Industry 4.0 technology: higher production volumes, lower costs, better and more timely control over quality. In particular, producers of complex goods like automobiles and electronics stand to make dramatic improvements to the bottom line using AI- and IoT-driven automation.


But bringing these technologies to sophisticated manufacturing operations poses significant challenges. How do you reduce the bandwidth requirements and latency of transmitting terabytes (or even petabytes) of video data to the cloud to assess parts quality? Conversely, how do you create and deploy specialized AI models to hundreds or thousands of machines and devices on manufacturing floors? How do you scale and maintain these deployments across multiple facilities around the world? And how do you do all of this cost effectively?


The IBM Systems Manufacturing team faces these same challenges, and recently took them on.


The IBM Systems division runs factories around the globe that annually produce hundreds of thousands of today’s most advanced and powerful computers. Each individual mainframe, server, or data-storage unit is a complex and intricate creation requiring hundreds of manufacturing steps. Ensuring that each system meets stringent quality standards requires intensive visual inspection at numerous points. Performed manually, some of these inspections took several minutes—adding up to very significant time and cost when multiplied by total production volumes. Further, many inspection items are tiny and difficult to scrutinize by eyesight. Missing a physically miniscule defect can create large rework and cost implications that can extend into the field.  


With the goal of improving the efficiency and accuracy of quality inspections, the Systems team put some of IBM’s own technology to work in a prototype implementation that shows how a manufacturer can bring Industry 4.0 to the real world of the production floor.

Automated inspection at the edge

The Systems manufacturing team developed an AI-powered, automated inspection solution that runs at the edge for maximum efficiency. The solution combines two of IBM’s leading AI and edge solutions, IBM Maximo Visual Inspection and IBM Edge Application Manager software.


The AI in the solution comes from Maximo Visual Inspection software, a computer vision solution that applies deep learning models to automatically detect quality defects in materials or assemblies with very high precision. The intuitive toolset within Maximo Visual Inspection enables the Systems team to train visual inspection models without reliance on machine-learning experts. Once deployed, Maximo Visual Inspection can refine the inspection models on its own based on operator feedback. The team uses the IBM Cloud® Object Storage service and IBM Elastic Storage® System 3000 to provide unified storage for video data and inspection models.


The Edge Application Manager software enables the Systems team to bring the Maximo Visual Inspection solution to the edge—a critical capability for scaling the AI to global dimensions.


The visual inspection models are compute-intensive, and each inspection step for a given system requires its own specialized model. Running these workloads in the cloud or even across network connections to on-premises servers could tax bandwidth and create processing latency—potentially negating the automation’s efficiencies.


Using Edge Application Manager, the team deploys the inspection models to NVIDIA Jetson TX2 modules—edge devices that integrate with inspection cameras and actuators for inferencing. The NVIDIA devices run the models and process the visual data on the spot, preserving network bandwidth and producing inspection results rapidly. The team is also testing IBM Power® System IC922 servers as edge nodes in addition to the NVIDIA devices.


By providing central deployment control, Edge Application Manager lets the Systems team avoid weeks or months of local, manual installs to each edge node. From a single console, the team can deploy the inspection models to the right devices, at all of its sites around the world, then autonomically maintain and update the models and monitor the health of the devices dynamically and securely, at the speed of business.

Efficiency gains and cost savings

The Systems team now employs the visual inspection solution at the edge at plants in Canada, Hungary, Mexico and the US. The solution’s impact is dramatic.


A 5x increase in inspection efficiency and 20% reduction in false-positive defect detections help maintain the highest levels of quality and production throughput. As a result, the Systems team estimates multimillion-dollar cost savings across its product lines.


The ability to centrally and automatically deploy and manage the edge solution, via Edge Application Manager, provides several additional benefits:

  • 20% cost savings on software maintenance
  • Faster time to value
  • Minimized disruptions, supporting 24x7 production
  • Scalability to deploy the visual inspection solution to more operations, multiplying the benefits

IBM Systems Manufacturing

The IBM Systems team produces some of the world's most advanced mainframes and servers, including IBM Z®, IBM Power® Systems, IBM LinuxONE and IBM Storage systems. The team operates manufacturing facilities around the globe and produces hundreds of thousands of systems every year.