Overhead Crane Cloud-Edge Control System Architecture
Core Specifications
The overhead crane cloud-edge collaborative control system uses a three-tier architecture: the field device layer (crane PLC + sensors + AI cameras), the edge gateway layer (Jetson Orin NX edge computing box + industrial IoT gateway), and the cloud platform layer (AI training + time-series database + remote operation & maintenance). The edge tier handles real-time AI inference (5–12 ms) and data preprocessing, while the cloud tier manages model training, data analytics, and remote management. The three tiers communicate via Profinet (50–100 ms cycle), OPC UA/MQTT (cloud upload), and secure encrypted channels. Deployed across 30+ projects in the steel, cement, and automotive industries.
The overhead crane cloud-edge collaborative control system is the core technology pillar of Kelude's "Smart Crane" series. By deploying an edge computing box locally on each crane, real-time inference functions such as AI vision anti-collision and hook tracking are executed on-site. An industrial IoT gateway then reliably uploads crane operating data to the cloud platform for centralized analysis and predictive maintenance. Compared with purely local solutions (limited compute power, no global optimization) and purely cloud-based solutions (high latency, network dependency), cloud-edge collaboration strikes the optimal balance between real-time responsiveness (5–50 ms at the edge) and global visibility (cloud-wide analytics). This article provides a complete overview of the cloud-edge architecture design, core technology selection, and typical deployment scenarios, with links to three technical deep-dives covering each subsystem.
Three-Tier Architecture Breakdown and Layer Responsibilities
Tier 1: Field Device Layer.The crane PLC (S7-1200/1500) serves as the crane's "neural hub," controlling the hoist, crane bridge, and trolley VFDs (G120/SINAMICS) via Profinet IRT. It collects vibration, temperature, and current sensor data (via AI module sampling) and exchanges AI inference results with the edge computing box through a Profinet slave interface. AI cameras (IMX296 global shutter, 2 MP, 120 dB wide dynamic range) connect to the edge box via USB3.0 or MIPI CSI. Inter-layer communication: device-to-edge uses Profinet IRT (50 ms cycle); actuation and sensing are handled directly by hardware I/O (<1 ms).
Tier 2: Edge Gateway Layer.This is the core of the cloud-edge architecture, comprising two key devices: the edge computing box (Jetson Orin NX, 100 TOPS) handles AI inference—running YOLOv8n object detection (5–8 ms/frame), ByteTrack multi-object tracking (2–3 ms), and FFT spectrum analysis (vibration feature extraction). The industrial IoT gateway aggregates data and forwards it to the cloud—collecting the PLC's DB_Upload data block (64 bytes/100 ms) via Profinet and uploading to the cloud platform over MQTT/OPC UA, while maintaining a 7-day local SQLite cache (with automatic failover during network outages). Inter-layer communication: edge-to-cloud uses MQTT over 4G/5G (100 ms–5 s cycle, tiered by data type).
Tier 3: Cloud Platform Layer.Deployed on Alibaba Cloud ECS or Huawei Cloud IoT, the core components include: a time-series database (TDengine/InfluxDB) for historical data storage (write throughput ≥2 million points/s); an AI inference server (1D-CNN + LSTM models for health-state classification and remaining useful life prediction); a remote operation & maintenance web/APP interface (Grafana dashboards + work-order system + video inspection); and OTA upgrade management (firmware chunked delivery + SHA-256 verification + rollback mechanism). Inter-layer communication: HTTPS REST API (front-end dashboards), MQTT (real-time data push), and WebSocket (video stream).
Three Core Technology Pillars with Linked Deep-Dive Articles
Each of the three core technology pillars of the cloud-edge collaborative control system carries its own engineering complexity and technical depth. The dedicated articles are linked below:
① Edge Computing Box Selection & AI Inference Deployment—covering selection comparison across Jetson Orin NX/Nano/RK3588 hardware, YOLOv8n model training with TensorRT INT8 quantization, ByteTrack multi-object tracking, Profinet communication setup between the edge box and PLC, and deployment validation metrics.
Read more: Edge Computing Box Selection and AI Inference Deployment for Overhead Cranes
② Crane PLC Data Upload Protocols & Gateway Configuration—covering OPC UA vs. MQTT protocol selection, industrial IoT gateway hardware (IoT2040/AR650/Kelude custom gateway), data point table design with TIA Portal DB block configuration, security policies (network isolation/TLS encryption/identity authentication), and cloud upload acceptance standards.
Read more: Crane PLC Data Upload Protocols and OPC UA/MQTT Gateway Configuration
③ Cloud-Edge Predictive Maintenance & Remote Operation for Overhead Cranes—covering sensor configuration and data acquisition, edge-side FFT spectrum analysis with time-domain feature extraction, cloud-side 1D-CNN health-state classification with LSTM remaining life prediction, work-order workflows with APP push notifications, and remote monitoring/video inspection/OTA firmware upgrades.
Read more: Cloud-Edge Predictive Maintenance and Remote Operation & Maintenance for Overhead Cranes
Standard Configuration Packages and Quotation Reference
Kelude Heavy Industry: Overhead Cranes & Gantry Cranes Manufacturer
Kelude Heavy Industry is a professional manufacturer of overhead cranes and gantry cranes, integrating design, manufacturing, installation, and after-sales service. Our product range covers various types and tonnages, including single-girder cranes, double-girder cranes, bridge cranes, gantry cranes, explosion-proof cranes, and European-style cranes. We provide comprehensive solutions for a wide range of industrial applications.
Deployment Case Studies & Measured Results
Steel Plant Continuous Casting Bay (Standard + Basic mixed) — 12 overhead cranes: 8 equipped with the Standard edition (AI vision anti-collision + cloud data), 4 with the Basic edition (cloud data + Remote Monitoring only). After deployment, collision warning latency dropped from 3–5 seconds to 50 ms, and collision risk events fell from 23 to 0.4 per day. Operator satisfaction rose from 62 to 91 points. Annual unplanned downtime dropped from 8 to 3 events (the Standard edition's predictive maintenance system identified 2 bearing fault precursors, which were replaced during scheduled maintenance, avoiding unplanned downtime).
Cement Plant (Flagship edition) — All 6 overhead cranes equipped with the Flagship edition. In the 4th month after deployment, the predictive maintenance system flagged severe degradation (Label 3) of the high-speed shaft bearing in the hoisting gearbox of a 32t crane. On-site inspection confirmed cracks in the bearing cage, and replacement was completed during a scheduled maintenance window (6 hours). Had the fault gone undetected, a bearing seizure would likely have caused a hoisting mechanism failure within 2–3 weeks, resulting in at least 2 days of unplanned downtime (based on the production line's output of 1,000 metric tons of cement per day, an estimated production loss of approximately $44,400). The loss avoided by this single early warning exceeded the total investment in the system.
Key Performance Indicators: Before vs. After Deployment
| KPIIndicator | Before Deployment | After Deployment | Improvement Rate |
|---|---|---|---|
| Collision Warninglatency | 3~5Second(Manual Visual Inspection) | ≤50ms(AIVision) | Improvement60~100Times |
| Daily Average Collision Risk Events | 23Times | 0.4Times | Reduction98% |
| Yearunplanned downtime | 8Times | 3Times | Reduction62% |
| Operator Satisfaction | 62Points | 91Points | Improvement47% |
| spare parts inventory Cost | Baseline(100%) | 55% | Reduction45% |
| Cloud Data Upload Completeness Rate | — | ≥99.9% | — |
FAQ: Cloud-Edge Collaboration System Integration & Upgrades
Q: Can the cloud-edge collaboration system integrate with AGV/RGV dispatching systems?
A: Yes. The AI vision and predictive maintenance data from the cloud-edge overhead crane system can be output to the upper-level dispatching system via OPC UA. The crane's current operating mode and fault prediction results serve as inputs for dispatching decisions—for instance, when the predictive maintenance system flags a crane's health status as "severely degraded," the dispatching system automatically lowers that crane's task priority and reassigns heavy-load tasks to healthier units. Kelude's crane dispatching system (see Overhead Crane & AGV/RGV Collaborative Operation Solution) comes with pre-configured cloud-edge data interfaces and supports health-weighted task allocation.
Q: What are the network requirements? What if 4G signal is weak?
A: The basic version only requires a stable 4G connection (RSRP≥-105dBm), with an average bandwidth demand of ≤50KB/s for 20 overhead cranes. If 4G coverage is insufficient, we recommend installing an outdoor high-gain antenna (8~12dBi gain, approx. $45) or an industrial CPE device (approx. $220, dual-SIM aggregation). In extreme cases (underground workshops, signal dead zones), a local edge server can be deployed on-site as a localized alternative to the cloud platform—edge box data is written directly to the local server without relying on public networks. Kelude will recommend the optimal solution based on a site network survey.
Q: Do we need dedicated IT staff to maintain the system after deployment?
A: No. The cloud-edge collaboration system is designed as "plug-and-play"—the edge box and gateway come pre-loaded with all software and configuration files, and start working as soon as they're powered on. The cloud platform is centrally managed by Kelude's operations team (including server monitoring, model updates, and security patches). Your only responsibility is keeping the cranes running and the 4G connection active. In daily use: operators check equipment status via the mobile app, and maintenance managers confirm and dispatch work orders through the same app. Kelude provides a 7×12-hour technical support hotline, with a 2-hour response time for major failures.
Q: Can existing Basic or Standard version users easily upgrade to the Flagship version?
A: Yes, the upgrade path is fully supported and straightforward. For Basic to Standard: add a Jetson Orin NX edge box and AI camera, then connect one additional Ethernet line from the existing gateway to the edge box—existing cloud data upload functionality remains unaffected. For Standard to Flagship: add the vibration/temperature/current sensor kit, connect sensors to the edge box's GPIO/SPI interfaces (pre-reserved for this purpose), and activate the predictive maintenance module on the cloud. All upgrades support remote software configuration, requiring no additional on-site work (except hardware installation). Kelude provides lifetime upgrade path support for all customers.
