AI Enters the Shop Floor Physical AI in Manufacturing: Eight Examples

Related Vendors

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

SIMTOS (Seoul International Machine Tool Show)

Physical AI connects the digital with the real world. The systems are capable of responding to new and unfamiliar situations and learning from their experiences. NTT Data, a global leader in AI, digital business, and technology services, identifies eight areas where Physical AI delivers the greatest added value.

The history of manufacturing automation spans many decades, but with AI and robotics, a completely new chapter is currently being written, as it is no longer limited to highly standardized processes. Even complex tasks and workflows with high variability are becoming increasingly automatable, and unforeseen events no longer disrupt production. The key is Physical AI, which enables intelligent control of physical systems in real time, thereby revolutionizing their adaptability.

"Intelligent machines that not only think but also act are the future of manufacturing," emphasizes Oliver Köth, Managing Director of Technology & Innovation at NTT Data DACH. "They can help manufacturing companies mitigate the labor shortage, optimize complex manufacturing processes, and even automate workflows that previously seemed unautomatable. Those who neglect this topic are unlikely to remain competitive in the long term."

Interview With Zhaopeng Chen, CEO of Agile Robots

Physical AI is Revolutionizing Industrial Robotics

Particularly Promising Application Areas for Physical AI

• 1. Intralogistics for Small and Micro Series: Products manufactured in small quantities or customized have traditionally posed a logistical challenge. However, with Physical AI, it is now possible to deliver the right parts to the right place at the right time. Autonomous Mobile Robots (AMRs) independently pick them in the warehouse and transport them to the machines. They select the best possible route, avoid obstacles, and prioritize short-notice rush orders if necessary. They also handle the transport of semi-finished goods between individual production stations, preventing idle times or backlogs of unfinished products. This reduces lead times and increases on-time delivery, while companies also need to maintain smaller inventory buffers, as demonstrated by various projects that NTT Data has already implemented in the automotive and mechanical engineering sectors. Since the AMRs continuously learn from feedback in the ongoing process (Closed Loop), they do not require regular retraining based on recorded data.
• 2. Individual Machine Control: Small and micro series require many individual adjustments to machines, but Physical AI can make these dynamically. Depending on the product and its requirements, the right materials and components are ordered and then processed correctly. For example, a gripper recognizes what is in front of the machine and grips it precisely—without toppling over or damaging parts, even if they are new to it. During assembly, the AI selects the appropriate tools and parameters, such as the desired paint and suitable nozzle for painting or adjusting welding current, welding time, and electrode force according to the material type and thickness. This increases the flexibility of production variants without the need for manual parameterization—setup times are significantly reduced, and new products can be produced error-free more quickly (First Time Right). In brownfield environments with existing PLC and MES landscapes (Programmable Logic Controllers / Manufacturing Execution System), machine controls based on Physical AI have already proven themselves many times.
• 3. Maintenance Optimization: Predictive Maintenance is not new but gains reliability with Physical AI. Instead of merely analyzing statistical data on past disruptions and failures to optimize maintenance cycles, AI can now also incorporate sensor data to reliably predict wear and defects. It is important not to rely on individual sensor data such as vibrations, sounds, temperatures, or pressure but to create a comprehensive picture through their integration—known as sensor fusion. If necessary, the AI can also adjust machine parameters or stop the machine in time to prevent faulty batches or major damage. It autonomously schedules maintenance windows to optimally utilize downtime and replace parts neither too early nor too late.
• 4. Automated Quality Control: Systems equipped with cameras and AI can significantly enhance quality control in manufacturing. They detect even the smallest defects, such as microscopic cracks in materials and coatings or slight deformations and discolorations on surfaces that indicate flaws. They inspect weld seams, fastenings, gaps, and dimensions—faster and more accurately than humans ever could. Products that do not meet the specified quality standards are automatically sorted out or sent to the appropriate machines for rework. Thanks to a combination of image processing and machine learning, these systems deliver significantly better results than rule-based vision systems. Additionally, they can be directly integrated into the production process to inspect all components or products rather than just random samples. This reduces scrap rates and customer complaints.
• 5. Self-Optimizing Machines: Machines and tools that optimize themselves represent a combination of the aforementioned use cases. They utilize extensive sensors not only to monitor the results of their work but also to assess the current tool condition—for instance, detecting wear, fouling, imbalance, or a clogged nozzle. Based on this, manufacturing parameters are adjusted to ensure consistently high production quality despite wear or contamination and, if necessary, initiate actions such as cleaning, lubrication, recalibration, or part replacement. A similar approach can be used to optimize the ramp-up phase for manufacturing new products: test products are meticulously inspected to identify deviations and gradually adjust production parameters. Targeted parameter changes help to better understand causes and effects, allowing the entire process to be refined until the production result meets the specifications.
• 6. Autonomous Monitoring and Inspection: Independently operating drones and robots can monitor both the factory premises and production areas—either on regular inspection routes or when surveillance cameras and other sensors report an irregularity, such as unknown individuals, open doors, blocked escape routes, or smoke and fluid leaks. They also cover blind spots and reach areas that are inaccessible or hazardous for humans. By combining camera images with infrared and LiDAR, they can achieve highly detailed object and event differentiation, even in low-light conditions. Depending on what the sensors detect, appropriate measures are initiated, such as notifying security personnel, shutting down machines, closing valves, or activating ventilation systems.
• 7. Human-Machine Cooperation: Physical AI enhances collaboration between humans and machines. They can operate in the same workspace without safety cages or fences. Robots, for example, recognize when humans cross their path or enter their action radius and adjust their movements to avoid causing injuries. Furthermore, robots can detect the posture, movements, and gestures of their human colleagues and provide individualized support. Depending on the work situation, they lift heavy parts and hold them in a position suitable for ergonomically friendly processing, hand over the required tools, deliver components just in time, or transport assembled parts away.
• 8. Humanoid Robots: Robots whose form and movements are modeled after the human body can take on complex tasks in manufacturing that are usually performed by humans. At present, they are still mostly limited to simple activities, such as sorting and transporting parts, but development is advancing rapidly. Their use is particularly attractive in situations where there is a shortage of skilled workers or where boring, repetitive, physically demanding, or hazardous tasks need to be performed. Thanks to their arms and legs, humanoid robots can move freely and assist in a wide range of work processes—around the clock, without fatigue. Even autonomous factories, where autonomous robots and other intelligent systems fully organize themselves, now seem possible.

What Decision-Makers Should Consider

Physical AI requires not only extensive AI expertise but also seamless IT/OT integrations, as intelligent systems must, among other things, retrieve data from MES and access PLC and SCADA systems (Supervisory Control and Data Acquisition). Extensive sensor technology and high data quality are additional critical success factors, as the AI needs a comprehensive and accurate picture of processes and the environment. Companies should initially pilot individual use cases intensively and subsequently standardize and scale the automated processes. Governance and IT security must be considered from the outset—otherwise, compliance and security risks may arise, necessitating costly rework.

When carefully implemented, Physical AI projects deliver significant added value. According to NTT Data's experience, typical results include a reduction in downtime by 20 to 40 percent, a 10 to 25 percent increase in Overall Equipment Effectiveness (OEE), a 15 to 30 percent reduction in scrap production, and significantly shorter ramp-up times for manufacturing new products.