Crane Safety AI
Event Gates Crane Safety AI
Event Gates Crane Safety AI is the popular overhead crane safety solution built with VIS. It works with cameras installed under the trolley of the over head crane. VIS processes the camera feed, runs Event Gates proprietary AI models to detect violations and interact with the crane to slow it down, stop or prevent from starting.
Features
Crane Safety AI is a VIS solution therefore it has all VIS features and functions where applicable:
- Disconnected solution, deployed to the edge devices on the crane trolley or bridge, does not need internet or corporate network access.
- Proprietary high performance person detection model Event Gates PF, specially trained for top down view in industrial environments.
- Built-in trustworthiness features: Heartbeat, Short Term Memory, Emergency Preset of IPIO.
- Logging features including system logs, event logs and optional emailing features.
- All flexibility of regions(ROI), including polygons and multiple regions.
- Built-in mute with IPIO.
- Input-based selection of RoI enables switching between preset inputs for different load sizes.
Risks assessment
Lift-off Hazards
Lifting operations involving heavy loads, such as body dies, carry significant risks from the moment the load is lifted off the ground. During lift-off, dynamic forces and changes in sling tension can cause the load to sway or move unpredictably. If personnel are present within the immediate vicinity of the load, this unexpected movement creates a high risk of crushing injuries, as workers may be caught between the moving load and fixed structures or the ground.
In busy industrial environments, where multiple operations occur simultaneously and visibility or communication may be limited, the likelihood of personnel inadvertently entering these hazardous zones increases. The combination of the load’s substantial mass, momentum, and unpredictable motion during lift-off demands strict spatial controls and operational discipline to prevent severe injury or fatality.
Object drops
In many industrial lifting scenarios, the transported load may consist of multiple components or unsecured parts that are not rigidly fixed to the main body. During transit, especially when the crane accelerates, decelerates, or changes direction, these parts can shift, slide, or fall if not properly secured. Such partial load drops pose significant hazards to personnel working nearby and may result in damage to equipment, tooling, or infrastructure. These incidents are particularly difficult to anticipate, as the main load may appear stable while internal or secondary elements are at risk of displacement.
This risk is especially pronounced in operations involving tools, dies, stacked materials, or custom fixtures where secondary elements are present. The appearance of a stable load can be misleading, as internal movement or external impact during transit may lead to partial drops. Such events are difficult to predict through visual inspection alone and are often exacerbated by limited visibility, time constraints, or assumptions about load integrity.
Sling placement
Slings are widely used in industrial lifting operations to secure and transport loads with overhead cranes. However, improper sling configurations present significant operational and safety risks. Incorrect placement—such as asymmetrical alignment, twisting, or slings positioned at non-optimal angles—can result in unstable lifting conditions, unintended load movement, or contact with fixed structures. These issues are often introduced during manual rigging and may not be visually detected during standard pre-lift checks.
A second critical risk arises during the unloading phase, where slings may become unintentionally entangled or remain attached to the load. If the crane continues operation while the sling is still engaged, it can lead to sudden shifts in load position or uncontrolled release, both of which pose a danger to nearby personnel and equipment. These failure modes are difficult to prevent through procedural controls alone and often require automated systems capable of detecting unsafe sling conditions in real time.
Solutions
Operator Safety
In this patented solution, two or four cameras are mounted beneath the trolley to monitor the area around the load, depending on the geometry and required coverage. VIS processes the video feeds and alerts the crane PLC when individuals are detected in hazardous proximity. While the use of a PLC is not mandatory, it is recommended for handling load cell data and crane movement. A load cell is also advised to distinguish between loaded and unloaded states and to enable proper logging in case of safety violations during lifting operations.
Sling Safety
Two additional cameras, positioned to view the slings from above at an appropriate angle, allow VIS to detect the angle and condition of the slings. The Safety AI Sling Safety add-on prevents the crane from lifting if the slings are not properly positioned—whether during heavy lifting or empty lifting. When used in conjunction with a load cell and PLC, the system can also detect sling entrapment, identified by the crane registering a load while performing what should be an empty lift.
The Sling Safety add-on is compatible with both chain and fiber slings. When used with wire rope slings, additional buoys are required to ensure proper detection.
Use Cases
Automotive Body Dies
In automotive stamping factories, large body panels are formed using high-tonnage presses and interchangeable dies. These dies, typically made of steel, are regularly transported between storage areas and press lines using overhead cranes. A single die set can weigh between 10 to 40 tons, depending on the vehicle component and press specifications. The overhead cranes used in these operations typically operate at heights ranging from 6 to 10 meters above the factory floor. The handling process involves lifting these heavy and bulky loads with slings or dedicated lifting fixtures, often in environments with frequent personnel movement and limited visibility. Given the load size, weight, and operating height, precise rigging and controlled movement are critical to ensuring safety and avoiding collisions, injuries, or production downtime.
Event Gates Crane Safety AI enhances operational safety in automotive stamping environments by preventing crane movement when personnel are detected in hazardous proximity to the load. The system continuously monitors the area beneath and around the suspended load using AI-powered vision technology, ensuring compliance with exclusion zone requirements during lifting and transit.
By reducing the likelihood of personnel being present under the load, the system helps mitigate the risk of crushing incidents, uncontrolled load sway, and potential object drops. This contributes to a safer and more controlled working environment without disrupting production efficiency.
For automotive OEMs that frequently handle car body dies, a four-camera configuration is recommended in conjunction with the Sling Safety add-on. The system should ideally be integrated with a PLC and a load cell to enable monitoring of load conditions and lifting states. To mitigate crushing hazards, lift-off—whether under load or empty—should be inhibited when personnel are detected in close proximity to the suspended load. Similarly, the system can be configured to prevent or decelerate landing operations when individuals are present near the drop zone. The Sling Safety add-on also ensures that lifting is blocked in cases of misaligned slings or when slings are unintentionally caught on the load.
Injection Molds
Injection molds—used in both plastic and aluminum molding processes—are frequently transferred between storage and injection machines using overhead cranes. These molds are typically large, heavy, and cubic in shape, making them challenging to maneuver, especially in confined machine bays. Depending on the mold size and crane height, a two-camera configuration may be sufficient to monitor the area around the load. However, for larger or more complex transfers, additional cameras may be considered to ensure adequate coverage.
A particular risk in mold transfer operations arises during mold installation into the injection press. This process typically requires manual attendance. The crane operator is often positioned close to the mold and does not have a clear line of sight to the opposite side. If another person enters the area—such as a technician assisting with mold alignment—they may not be visible to the operator, increasing the risk of crushing between the mold and fixed structures, or of the mold being inadvertently landed on a person’s feet.
To address this risk, the system should be configured to reduce crane speed when a person is detected near the load and to fully stop movement if a second individual is present. This use-case requires enabling a person-counting function within the Crane Safety AI system to distinguish between the operator and any additional personnel. The ability to dynamically respond to proximity and personnel count enhances safety during mold installation, where close collaboration is often unavoidable but must be tightly controlled.
Coils - Steel and Aluminum
Steel and aluminum coils are routinely handled using overhead cranes equipped with C-hooks or C-tongs in processing plants, service centers, and manufacturing facilities. These coils can weigh up to 50 tonnes, and their cylindrical yet compact form makes them suitable for two-camera monitoring configurations—depending on crane height, coil dimensions, and aisle width.
In typical operations, the lifting attachment (C-hook or C-tong) remains permanently mounted and contributes significantly to the total suspended weight. Therefore, the system operates in always-load mode, and a load cell is not required to distinguish between loaded and unloaded states.
Due to the high mass and potential instability of the coil during initial lift, there is a critical risk of crushing or impact if personnel are present near the load. To mitigate this, lifting is inhibited when personnel are detected in proximity to the coil. During horizontal movement and landing, the presence of individuals near the load should result in automated speed reduction to minimize the risk of injury in case of unintentional load shift, swing, or descent. This configuration supports continuous operation while maintaining personnel safety in high-throughput, heavy coil handling environments.
Pipes and Other Long Products
In facilities handling long products—such as steel pipes, beams, or structural profiles—overhead cranes are commonly used to transport loads of extended length and variable geometry. Due to the longitudinal size of these materials, a four-camera configuration is typically required to ensure full coverage of the hazardous area around the suspended load. In more complex layouts or for exceptionally long loads, custom deployments with additional cameras may be implemented to maintain effective personnel detection and zone monitoring.
These loads are often lifted using electromagnetic tongs or specially designed mechanical grabs capable of handling multiple products simultaneously. In such configurations, a load cell is generally not required, as the lifting device itself provides consistent load behavior and the system is operated in always-load mode. Certain lifting systems also support 90-degree rotation of the load, introducing additional movement planes that increase the risk to nearby personnel.
To adapt to this variability, the region of interest (ROI) for personnel detection can be dynamically adjusted using input signals from the crane or lifting device—such as tong position, load orientation, or operational mode. This enables the safety system to maintain accurate and context-aware protection throughout the handling cycle. Lifting should be disabled when personnel are detected within defined exclusion zones, and crane speed can be reduced during travel and set-down phases when proximity risks are elevated.
FAQ
Does Crane Safety AI require a PLC?
No, Crane Safety AI does not require a PLC to operate. It can function independently and even trigger actions[1] by connecting its outputs directly to the crane’s drivers. However, we strongly recommend using a PLC to give system integrators greater flexibility and control. This allows them to tailor the response logic to meet specific end-user requirements.
Who defines violations and actions
Violations and what actions should follow shall be defined by the end users. These decisions must be based on their risk analysis of the lifting operation. System integrators are responsible for implementing these requirements.
Is a load cell necessary for Crane Safety AI?
Not necessarily. It depends on how the end user wants the system to behave:
- If the crane should move without intervention when not carrying a load, then a load cell is required to determine the load status.
- If the crane uses a heavy attachment that already poses a risk, then it is effectively always on load, and a load cell isn’t needed.
- If the crane must slow down or stop whenever a person is in the region—regardless of load—a load cell is again not required.
In all cases, the load cell (if used) should be connected to the PLC, where decisions are made. If load status is also provided to Crane Safety AI via IPIO digital inputs, it can be used to differentiate violations that occur specifically when the crane is under load—such as lifting near people.
Is is possible for Crane Safety AI to detect the load?
No, Crane Safety AI does not detect the load itself. It focuses on detecting people and their presence in defined zones. Load status must be determined by external devices like load cells and integrated via the PLC or input modules.
Is it possible for AI to detect the size of the load and adjust safety regions accordingly?
While it's technically possible to estimate the size of visible objects from above[2], neither AI nor a human can determine how stacked items are interconnected. Therefore, it's not safe to rely on visual input alone to define the size of the lifted load to adjust regions dynamically. This must be known by the operator in advance, and the appropriate safety region must be predefined based on that knowledge. For that safety reasons, dynamic regions are not supported by Crane Safety AI, but it is possible to switch between predefined regions based on external input.
Can Event Gates Crane Safety AI also detect forklifts?
Yes. Event Gates Crane Safety AI uses our proprietary Person Forklift (PF) object detection model, which accurately detects people, forklifts, reach trucks, and similar vehicles. This capability helps address the serious risks forklifts pose under crane working areas—including the potential for collisions that may lead to derailment or crane failure.
Thanks to VIS’s multi-region feature, operators can define wider or separate regions for forklift detection, allowing the system to trigger crane reactions—such as alarms or halts—when a forklift enters the operation zone.