Siemens and P&G Expand AI-Based Quality Inspection Across Global Manufacturing Operations
Time:2026-09-23 Browse: 0
Siemens and Procter & Gamble are expanding an artificial intelligence-based quality inspection solution across P&G manufacturing operations worldwide, marking another significant development in the integration of AI, machine vision, edge computing, and industrial automation.
The system, known as the Visual Inspection Cockpit, is designed to inspect products in real time while production lines are running at full speed.
Unlike traditional machine vision systems that may require extensive reconfiguration when products or materials change, the AI-based solution is designed to handle a wider range of variations.
The companies report that the system has reduced scrap rates by between 10% and 20%, depending on the product, while new deployments can be commissioned five to ten times faster than traditional customized vision systems.
For industrial automation engineers, the development demonstrates how AI is moving from experimental applications into production-level control and quality processes.
From Traditional Vision Inspection to Industrial AI
Machine vision has been used in factories for many years.
Cameras can inspect dimensions, colors, labels, positions, surface conditions, assembly results, and other characteristics.
However, conventional vision systems often depend on carefully configured rules.
If the lighting changes, packaging changes, materials behave differently, or a new product is introduced, engineers may need to adjust the inspection system.
This becomes especially challenging in high-speed consumer goods manufacturing.
Products may be flexible, textured, decorated, or slightly different from one production batch to another.
Traditional vision technology can still be highly effective, but maintaining inspection accuracy across large numbers of products and production conditions can require significant engineering resources.
AI-based vision provides another approach.
Instead of relying entirely on fixed rules, deep-learning models can learn patterns associated with acceptable and defective products.
This allows the inspection system to recognize more complex variations.
How the Visual Inspection Cockpit Works
The Visual Inspection Cockpit was developed jointly by Siemens and P&G.
The system combines P&G's deep-learning models with Siemens Industrial Edge technology and industrial computing hardware equipped with NVIDIA GPUs.
Inspection takes place close to the production equipment.
This edge-based architecture is important because manufacturing quality decisions often need to happen within milliseconds or seconds.
Sending every camera image to a remote cloud platform may introduce unnecessary latency and network requirements.
Processing data locally allows the inspection system to respond quickly.
If a defective product is identified, the system can communicate with the manufacturing control environment and trigger an action such as an alert or product rejection.
This creates a direct connection between AI and conventional industrial automation.
PLC Integration Is a Critical Part of the System
For PLC engineers, one of the most interesting aspects of the project is the integration between AI-based inspection and conventional control systems.
An AI model can identify a defect, but that information must then be converted into an industrial control action.
For example, a production line may be moving thousands of products per hour.
When the vision system identifies a defective item, the control system needs to know exactly which product is defective and when it reaches the rejection mechanism.
The PLC may then control a pneumatic actuator, servo mechanism, diverter, or other device to remove the product.
This requires precise synchronization between cameras, image processing, product tracking, encoders, sensors, PLC logic, and physical actuators.
It demonstrates why industrial AI cannot be treated as a standalone software application.
AI must ultimately operate within the deterministic environment of a production line.
Edge Computing Brings AI Closer to the Machine
The use of industrial edge computing is another important element.
Traditional industrial architectures often separated control systems from advanced computing platforms.
Today, edge computing allows more processing power to be placed closer to production equipment.
Industrial PCs can run AI models, analyze sensor information, process images, and exchange information with PLCs and other control devices.
This architecture can reduce dependence on centralized cloud infrastructure.
It can also improve response time and help keep sensitive production data inside the industrial environment.
For manufacturers with multiple plants, edge computing can provide a standardized platform for deploying industrial applications across different locations.
This is especially valuable when a successful application needs to be replicated across multiple production lines.
Scaling AI Across Manufacturing Plants
One of the most important aspects of the Siemens and P&G project is scalability.
Developing a successful AI inspection system for one production line is only the beginning.
Large manufacturers operate hundreds of production lines across multiple countries.
If every factory needs to develop its own AI system from scratch, the engineering effort can become enormous.
A standardized Industrial Edge application can potentially make deployment more repeatable.
The same basic architecture can be adapted for different products and inspection requirements.
This changes the economics of industrial AI.
Instead of treating every AI project as a unique engineering project, manufacturers can begin to treat AI applications as reusable industrial software components.
This is similar to the way PLC function blocks, automation libraries, and standardized engineering templates are reused across multiple projects.
Reducing Scrap Through Better Quality Control
Quality inspection has a direct relationship with manufacturing efficiency.
A defective product can result in wasted materials, energy, packaging, labor, and production capacity.
If defects are discovered only after a production batch has been completed, the potential amount of waste can be much larger.
Real-time inspection provides an opportunity to detect problems earlier.
For example, if a packaging process begins producing incorrectly positioned labels, an automated vision system can identify the problem while production is still running.
Operators can then investigate the cause.
The inspection system may also automatically reject affected products.
The reported 10% to 20% reduction in scrap varies by product and application, but it illustrates the potential economic value of combining AI with real-time quality control.
AI Does Not Replace Industrial Automation Engineering
The expansion of AI inspection should not be interpreted as the end of traditional automation engineering.
Instead, it adds another technology layer.
A complete AI inspection system still requires cameras, lighting, industrial computers, network infrastructure, PLC integration, sensors, actuators, mechanical equipment, safety systems, and maintenance procedures.
Engineers must also ensure that the AI model operates reliably under actual production conditions.
Training data needs to represent real products and real defects.
Lighting conditions must remain controlled.
Camera positioning must remain stable.
Communication failures need to be detected.
The PLC must respond correctly if the AI system becomes unavailable.
These are classic industrial automation engineering challenges.
The Importance of Industrial Data
AI-based manufacturing depends heavily on data quality.
A model can only perform well when it has appropriate training and validation data.
Manufacturers therefore need reliable processes for collecting images, identifying defects, labeling data, monitoring model performance, and updating models when products change.
This introduces a new responsibility for manufacturing engineering teams.
In traditional automation, engineers primarily configured control logic.
With industrial AI, teams may also need to manage datasets, AI models, model versions, edge applications, and performance metrics.
This creates a new intersection between automation engineering, data engineering, and artificial intelligence.
Connecting AI With Smart Manufacturing
The Siemens and P&G project also illustrates a larger trend toward connected manufacturing.
A quality inspection system can produce much more information than a simple pass-or-fail signal.
Inspection data can be analyzed over time to identify recurring problems.
If a particular defect becomes more frequent, engineers can investigate upstream process conditions.
The data may be combined with machine status, production parameters, temperature, pressure, speed, vibration, or maintenance information.
This can help manufacturers move from simple defect detection toward process optimization.
In the future, AI-based inspection could become part of a broader closed-loop manufacturing system in which quality data influences production parameters automatically.
What This Means for the Industrial Automation Market
The expansion of AI-based inspection has implications for the broader automation industry.
Demand may increase for industrial cameras, lighting systems, edge computers, GPUs, industrial Ethernet, PLCs, high-performance controllers, data platforms, and AI software.
System integrators will also need to develop new skills.
Traditional automation engineers may increasingly work alongside machine learning specialists and data engineers.
At the same time, AI specialists working in manufacturing need to understand industrial requirements such as cycle time, deterministic control, machine safety, PLC communication, and equipment lifecycle management.
The future factory will therefore require collaboration between several technical disciplines.
The Next Stage of AI in Manufacturing
The Siemens and P&G deployment demonstrates that industrial AI is becoming increasingly connected to real production processes.
The technology is no longer limited to experimental demonstrations.
AI can now participate directly in inspection, quality control, production monitoring, and automated rejection.
For manufacturers, the key question is not simply whether AI can identify defects.
The more important question is how AI can be integrated reliably into the complete automation architecture.
That architecture may include PLCs, DCS systems, industrial robots, sensors, machine vision, edge computers, industrial networks, MES platforms, and enterprise software.
As manufacturers continue investing in smart factories, this type of integrated approach is likely to become increasingly important.
The combination of AI, edge computing, machine vision, and industrial control represents another step toward highly connected manufacturing environments where production decisions can be made faster and supported by real-time operational data.
