Cost-effective and comprehensive inspection based on machine learning

Automating quality control with AI-based image processing requires a suitable network and automation concept. Phoenix Contact supports system evaluation and component selection, as shown in this example from an international automobile manufacturer.

The automobile manufacturer maintains high quality standards by inspecting and optimizing all car components. Automated quality inspection is crucial for cost-effective control, using sensors (like ultrasonic sensors or cameras) to detect defects, which are then evaluated and corrected by robots. In the body shop, weld quality is a major focus, as proper welding ensures vehicle stability and durability. Additionally, removing weld spatter is essential to prevent damage to cables laid later.

Previously, inspection involved manually swiveling car bodies and illuminating them with diffuse lighting, then removing weld spatter with a Dremel tool. To automate this, the manufacturer conducted a feasibility study using eight cameras for inspection and a robot-guided rotary tool to remove the weld spatter.

Flexible fields of application due to the generalization properties

Implementing automated weld spatter removal with cameras and robots requires additional software. Firstly, this software synchronizes cameras and lighting for brightness control. Next, it inspects images for weld spatter. Then, it converts the position into the robot’s coordinate system. Finally, it guides the robot to remove the spatter with its rotary tool.

In this system, industrial image processing detects weld spatter, implemented either rules-based or via machine learning (ML). The manufacturer chose ML for its flexibility with different component shapes. The pretrained ML model learns to detect weld spatter based on properties like shine, shadows, or shape, making it adaptable to various automotive parts and spatter appearances.

In the ML process, images of components with and without weld spatter are taken. The spatter is manually identified and described using coordinates or bounding boxes. This data trains the model, and the descriptions check recognition accuracy. The trained model is then tested on new images. Due to generalization, it can detect spatter on previously unprovided images, serving as an identifier for removal.

Joint evaluation of a coordinated automation and network concept

Besides the automated quality control solution, a coordinated automation concept is needed. This includes connecting and powering cameras and lighting, using a PLC to control the line and robots, execution units for ML processes, and a control station for human-machine interaction. Additionally, components must form an efficient, interference-free network. Phoenix Contact assisted in creating the concept and evaluating suitable ML hardware.

Proven systems from the Blomberg-based company, already in use at the automobile manufacturer, were utilized for automation, network connection, and control stations. For ML processes, the manufacturer provided pretrained models, which Phoenix Contact integrated into a machine learning runtime environment and tested on various industrial PCs (IPCs). CPU and RAM utilization were verified in endurance tests. The results helped select the most cost-effective hardware. Phoenix Contact also offers hardware, software, and services for other industries through its Digital Factory now campaign.

Future potential through virtualization, GPU extensions, and Time-Sensitive Networking

The chosen concept perfectly meets the automobile manufacturer’s current needs. Future ML technology developments will offer more optimization potential. Container virtualization will enable automation services and ML processes in IT or cloud environments. Model execution can be accelerated with industrial PCs using GPUs. Currently, automation and camera networks must be separate, but Time-Sensitive Networking (TSN) can combine them, reducing complexity. Phoenix Contact is working on TSN applications. Regular technology days and information exchanges help tailor these technologies to the automotive industry’s needs.

The automobile manufacturer already uses ML processes for automated quality control. Phoenix Contact supports evaluation, automation, and network concepts with its hardware, software, and services.

Read more about AI-based quality control in the body shop.

Fewer emissions with a higher overall equipment effectiveness

More countries are committing to net-zero emissions, aiming to limit global temperature rise to 1.5 °C. With the PLCnext Technology ecosystem and MLnext software solution, Phoenix Contact is helping tp reduce emissions.

The EU aims for net-zero greenhouse gas emissions by 2025, as part of the European Green Deal and the 2015 Paris Agreement. This transition to a climate-neutral society is both a challenge and an opportunity, involving all societal and economic sectors, from energy and industry to buildings, agriculture, and forestry.

Conversion of existing thermal processes along the entire production chain

For sustainable climate strategies to succeed, all areas of industrial facilities must be considered. Failure to do so will undermine efforts to reduce greenhouse gases and comply with environmental directives.

Energy-intensive industries must convert thermal processes to low-carbon methods urgently due to climate-driven geopolitical changes. This requires sustainable, minimally invasive technologies and secure digitalization platforms. Heating systems should use renewable energy, ensuring energy availability, product quality, and cost-effectiveness. Conventional heating concepts need rethinking for low-carbon heat generation.

Need for change through the use of hydrogen

The power-to-gas (PtG) principle, which converts electrical energy into chemical energy (gas) by means of water electrolysis, is crucial for modern gas and hydrogen networks. This creates new requirements for industrial furnaces.

Hydrogen and its mixtures lead to a reduction in the air requirements, a change in the calorific value, higher combustion speeds, higher flame temperatures, and an increase in nitrogen oxides (NO and NO2). This impacts the “machine,” including the melting tank, furnace, boiler, supply technology, burner, and especially the controller and automation system.

Requirements on the thermal processing technology

Integrating thermal processes into existing systems must be efficient and minimally invasive. Past digitalization often overlooked crucial mechanical equipment for fuel-air ratios and other key functions.

Future thermal processing technology requirements: • Operation on open, holistic ecosystems/platforms • Scalable integration of combustion and exhaust gas sensors • Functional safety and cybersecurity per IEC 62443 and NIS 2.0 • Introduction of ML and IIoT for monitoring and optimization • Increased agility with low/no-code apps for adapting fuel lines using platforms like Grafana

Machine learning for fuel and process optimization

By detecting unusual behavior patterns early, PLCnext Technology with MLnext software helps prevent malfunctions in combustion processes, enhancing overall equipment effectiveness (OEE), optimizing fuel-air ratios, and enabling intelligent cross-factory monitoring.

Phoenix Contact’s Digital Factory now initiative supports the All Electric Society principle, offering scalable, data-secure automation solutions via the PLCnext Technology ecosystem. In addition, these solutions facilitate easy integration into thermal processing machines worldwide, aiding manufacturers, system integrators, and operators in creating tailored products and systems.

Certification of functional and data security

The PLCnext Control product family, certified by TÜV Süd for OT Security (IEC 62443-4-2) and existing safety certification, sets new standards. The IEC 62443 standard emphasizes linking security and safety in automation solutions. With a focus on OT security and remote access, the PLCnext Control family complies with the upcoming Machinery Regulation.

Signal connection through intelligent isolators

Users must adapt fuel-air ratios to new fuels using parallel sensor systems for efficiency without impairing operation. Existing infrastructures should be converted to IEC 62443-compliant segmented networks to prevent data collisions or illegal access.

Both goals can be achieved in parallel. Field signals connect to PLCnext Control or edge devices using Mini Analog Pro signal isolators and the “PLCnext IIoT Framework” app. Phoenix Contact’s configurable thermal processing control cabinets process, normalize, and send signals in real time, saving space, hardware, and engineering time. The MLnext app can virtualize information from various sources on-premise or in the cloud (e.g., Proficloud.io).

Read more about PLCnext Technology.

Foundation for optimizing the entire value chain

Edge computing is closely tied to cloud computing, with the latter gaining importance in Industry 4.0 projects. Though relatively new, edge computing’s potential is exemplified by the EPC 1502 edge device.

A user needs an Internet connection to link systems to a cloud for data evaluation and storage. However, for applications requiring quick data processing, public cloud solutions are unsuitable due to Internet latency. Users are increasingly using Edge computing for these smart applications. Phoenix Contact’s edge devices, based on the PLCnext Technology ecosystem, combine robust industrial PCs with an open automation platform. This approach enables intelligent IoT edge concepts and optimizes the entire value chain by processing and analyzing large data volumes locally.

Easy to integrate box for collecting local data

Edge devices have broad applications, where data needs to be analyzed or is already being analyzed in the cloud. When developing edge applications, users often ask how to centrally accumulate data. Creating autonomous vehicles, traffic light systems, or intelligent thermostats that benefit from edge computing can be complex. The edge device must be flexible and able to collect large amounts of data from various sensors and manufacturers. Additionally, field devices use different industry protocols, and much of the data is analog, making accumulation challenging.

The tailor-made Edge Collection Box from Phoenix Contact is ideal for collecting local data. It gathers digital, analog, and temperature signals via connected sensors and integrates easily into existing production environments. Furthermore, Data collection takes place in parallel, preserving the CE marking of machines. The main component, the EPC 1502 PLCnext Control with an IIoT framework, facilitates easy collection of energy and process data using industry-standard communication protocols.

Secure Data Box as a secure interface to higher-level IT systems

Once data is collected and analyzed locally via the edge device, it often serves as an intermediate layer to the cloud. Transferring summarized data, evaluated alarms, and analyses securely to the cloud can be challenging. Moreover, Phoenix Contact’s Secure Data Box, used with the Data Collection Box, provides a secure interface between the production network and higher-level IT networks, external service providers, and cloud systems. The integrated FL mGuard security router’s firewall blocks unauthorized access, allowing only necessary data traffic. Via the mGuard Secure Cloud app, one can access the EPC 1502 edge device for remote maintenance information.

Reliable operation of automated guided vehicle systems

Additionally, using an edge device in an automated guided vehicle system (AGVS) bridges the gap between classic control functions and AGVS navigation. It offers a plug-and-play interface for the Robot Operating System (ROS) and supports high-level language programming, the ROS interface, and Docker software, allowing users to implement their own approaches. The PLCnext Technology online store provides apps for sending firmware and application updates to all controllers.

Furthermore, the edge device provides a secure entry point for remote maintenance and displays collected data via dashboards. AI algorithms analyze data to detect anomalies, enabling predictive maintenance and minimizing downtime. With atypical interfaces, data can be forwarded via WLAN, saving on cables. The large internal storage allows for long-term local data storage. Edge devices are essential for the reliable operation, efficiency, and flexibility of automated guided vehicle systems.

Wide range of possible uses in large and small applications

There are various applications for Edge computing. Ready-to-install edge boxes simplify implementing large systems that collect data using different protocols. Moreover, existing systems can be easily converted to edge computing with integration-capable solutions installed outside machines, retaining their CE marking. In smaller systems, edge devices must support many protocols and apps to provide added value with simple installation.

Read more about edge computing and PLCnext Technology

From hand tools to automatic tools

Before making an electrical connection, conductors need to be stripped a process that varies depending on the type and characteristics of the conductor. Stripping conductors is challenging due to the variety of types and their unique characteristics, with high demands on reliability and efficiency. Stripping involves removing the insulation to a certain length, as determined by the geometry of the terminal or connector. In addition, it can include removing the outer sheath of multi-core cables. Ensuring a safe electrical connection means avoiding damage to the litz wires, single-core wires, or braided shields. The process must conform to DIN IEC 60352-2 standards, which specify the requirements for a perfect stripping result. Depending on which tool is used, clamping the conductor can lead to pressure points and thus to discoloration of the insulation.

To avoid damage to litz wires, the stripping process involves two steps: “cutting into” and “pulling off” the insulation. First, the insulation is cut to a clearance of about 0.1 mm to 0.3 mm from the wires. Then, the uncut area of the insulation is pulled off to complete the removal.

Non-adjustable hand tools

Stripping can be performed using hand or automatic tools. Hand tools are categorized into non-adjustable, adjustable, and automatically adjustable types. Non-adjustable tools, like the Wirefox-MP VDE multifunctional tool, are the simplest and can handle a wide range of conductor sizes (0.2 to 10 mm²), allowing quick processing of different conductors. However, these tools require practice for perfect results and are not suitable when high level of process reliability is needed due to their lack of adjustability.

Adjustable hand tools

Adjustable tools enhance process reliability and require minimal practice. Tools with a rotating blade, like the Wirefox-D 40, are suitable for sheath removal and stripping in cross-sections over 10 mm², ensuring a uniform incision depth. For smaller cross-sections, non-rotating cutting edges are used, increasing the pull-off proportion.

While tools like the Wirefox-D 40 have broad applications, specific tools like the Wirefox D-CX series are designed for multi-stage stripping of coaxial cables. These tools are selected based on the connector type to achieve the required stripping lengths. However, they are not ideal for frequent material changes due to the time needed to set them up.

Automatically adjustable hand tools

Automatically adjustable hand tools offer the combined benefits of non-adjustable and adjustable tools, providing fast material changes and high process reliability. They use different blade geometries for various applications: flat blades for PVC cables up to 10 mm², half-round blades for cross-sections up to 16 mm² to minimize force required, and V-blades for hard insulation materials like PTFE and soft rubber. Stripping lengths can be set with an adjustable stop, and blades can be easily replaced without tools, ensuring long service life.

Automatically adjustable hand tools are also available for special applications, such as stripping SAC cables, AS-Interface flat cables, and flat-ribbon cables. These tools are recommended wherever applicable due to their versatility and efficiency.

Manually adjustable automatic tools

In series production, where quantities and quality demands are high, hand tools are reaching their limits. No matter how ergonomic the hand tools are, fatigue is inevitable. This is where automatic tools, like the entry-level electric WF 1000 stripping device, become essential. Once parameters such as cross-section, stripping length, and pull-off length are set, this device can quickly process many conductors without causing fatigue. However, each material change still requires some time to adjust the stripping parameters. This machine is especially suitable for processing identical materials efficiently.

Automatically adjustable automatic tools

The new E.Fox S 10 automatic stripping device is optimal for production processes that require frequent material or parameter changes. Its electrical adjustment of stripping parameters allows for swift material changes. The device is user-friendly, featuring an intuitive touch display interface. Users can save and recall favorite stripping parameters using a barcode scanner, significantly boosting efficiency.

The E.Fox S 10 automatic stripping device reduces the potential for human error compared to manual settings. It can function as a stand-alone device or be integrated with the Clipx Wire assist worker assistance software via an Ethernet interface. This allows for consistent use of production data and reliable processing of large, multi-version orders. When combined with other automatic tools and printing systems controlled by the same software, an efficient and customized worker assistance system for wire preparation can be created.

Read more about Automatic stripping devices.