Industry Background and Pain Points
Precision die-cast parts such as automotive engine blocks, transmission housings, and steering knuckles have complex casting processes that easily produce surface defects including sand holes, porosity, scratches, burrs, missing material, and cold shuts. These defects are distributed across multiple surfaces of the workpiece, making inspection difficult. Traditional inspection solutions use multiple independent inspection devices arranged in series along the production line. Workpieces pass through each inspection surface sequentially, which not only consumes large amounts of production line space but also makes signal synchronization between multiple devices difficult, easily leading to missed inspections and duplicate determinations. At the same time, casting workshops commonly face high ambient temperatures, heavy dust, and oil contamination. Ordinary commercial computers and consumer-grade GPU devices have high failure rates and short lifespans in such environments, making it difficult to meet 7×24 continuous production requirements.
EA Series Product Functions and Application Solution
This solution adopts EA230Pro edge AI computer as the inspection host, paired with four high-resolution GigE industrial cameras arranged in a surround layout. When the workpiece reaches the inspection station, image acquisition and defect recognition for all four sides are completed in a single pass—truly achieving four-surface full inspection at a single station. The EA230Pro’s hardware synchronization trigger mechanism ensures that the four cameras expose synchronously within a millisecond-level time window, avoiding inspection errors caused by slight workpiece displacement. Trigger modes support photoelectric sensor trigger, encoder trigger, PLC command trigger, and others, flexibly adapting to different production line control logic.
The device body uses a full aluminum alloy fanless reinforced chassis, relying on natural heat dissipation through the enclosure. There are no rotating parts inside, fundamentally eliminating the risk of dust and oil entering the device. The operating temperature range covers -20°C to +75°C, enabling long-term stable operation in the high-temperature summer environment of casting workshops. Power input supports a 12V–24V wide voltage range with overcurrent, overvoltage, and reverse-connection protection, adapting to factory grid fluctuation environments. The entire unit has passed CE, FCC certification, and EMC Level 3 electromagnetic compatibility testing, capable of resisting strong electromagnetic interference generated by welding machines, lifting equipment, and other machinery in the workshop.
For IO interfaces, the EA230Pro communicates in real time with the production line PLC through isolated RS485 interfaces. After inspection is completed, it immediately outputs OK/NG determination results and defect category codes to the PLC, working with the rejection mechanism to complete automatic diversion of nonconforming products. The full process of inspection–determination–rejection has a closed-loop delay of less than 50 ms. To address the challenge of low-contrast defect detection on metal surfaces, Beilai Technology provides dedicated image preprocessing operators and defect enhancement algorithms. Combined with deep learning models, these can effectively identify weak scratches and micro sand holes at the 0.1 mm level, with a comprehensive detection rate above 98%. Customers can also optionally add RTD and thermocouple acquisition modules through Y-board expansion to synchronously monitor workpiece temperature data, enabling multi-source data fusion analysis of visual inspection and process parameters.
Customer Benefits and Deployment Results
A well-known automotive Tier 1 supplier introduced this solution for transmission housing appearance inspection. A single workstation can complete four-surface full inspection of workpieces, saving 70% of floor space compared with the original series-type solution and making the production line layout more compact. The inspection takt stably reaches 40 pieces per minute—more than 5× the efficiency of manual visual inspection—and a single production line can reduce 6 quality inspectors. Since the equipment has been in operation for one year, no hardware failures have occurred in the high-temperature, dusty casting workshop environment, and the mean time between failures far exceeds customer expectations. The system supports defect type statistics and distribution heat map analysis, helping the process department identify mold locations with high incidences of sand holes. After targeted optimization of casting parameters, overall product yield increased by 3.2 percentage points, delivering significant economic benefits.
