| With the advancement of global science and technology,the replacement of electronic equipment is also more frequent,and the circuit board plays an irreplaceable role as a core component.The popularity of high-tech products such as smartphones and digital cameras has also brought challenges to the portability and reliability of circuit boards.Flexible circuit boards are the best solution because of their bendability and lightness.However,due to the complexity and fragility of the flexible circuit board manufacturing process,defects can occur during the production process.At present,most manufacturers use manual visual inspection methods for defect detection,but the detection efficiency is low and the accuracy cannot be guaranteed.Using artificial intelligence methods to detect flexible circuit board defects is one of the effective methods to improve detection efficiency and accuracy.However,if the detection system adopts the cloud computing solution,the defect detection task needs to be uploaded to the cloud computing center,and the cloud computing center will send back the results after the calculation is completed,which will cause a large delay.This paper uses the deep learning-based machine vision defect detection model of the project team to detect defects on flexible circuit boards.Using the edge computing method,the data transmission delay is reduced by offloading the computing tasks to the edge nodes deployed on the pipeline.The main work of the paper is as follows:1.Based on Docker container,a computing task offloading strategy aiming at minimum comprehensive delay is proposed.In the circuit board defect detection scenario,multiple circuit board defect detection tasks will be generated at the same time.How to choose the defect detection computing task offload scheduling strategy in the face of many edge computing nodes will have a key impact on the comprehensive delay.Aiming at the offload scheduling scenario of circuit board defect detection computing tasks,this paper proposes a computing task offloading strategy with the goal of minimum comprehensive delay based on Docker containers.In this offload scheduling strategy,after the collector generates a new computing task,the offload decision module in the collector first considers offloading computing tasks to the edge node closest to the collector.If the remaining computing resources and offloading execution delay of the edge node closest to the collector cannot meet the requirements,the collector offloading decision module offloads the computing task to the computing task scheduling module,and the computing task scheduling module obtains the calculation task according to the attributes of the computing task and the state of the edge node.Unload the delay matrix,and finally take the minimum delay as the greedy goal to find the minimum delay unloading decision,and complete the computing task scheduling.Through simulation,the computing task offloading strategy proposed in this paper is compared with the default computing task offloading strategy of k8s and the default computing task offloading strategy to the edge node closest to the collector,and it is verified that the proposed computing task offloading strategy can significantly improve Calculate the comprehensive delay of task execution.2.The circuit board collector,edge computing node cluster,and edge node manager are built on the virtual machine,and the circuit board defect detection computing task offloading strategy proposed in this paper is deployed.Further,in order to visualize the defect detection results,this paper builds a detection result visualization system on Alibaba Cloud based on the circuit board defect detection data visualization scene to display the circuit board defect detection results.The visualization system has three main functions,which are based on RBAC account authority system,circuit board defect detection data visualization and defect sample management library,which can well meet the needs of circuit board defect detection scenarios. |
PDF Full Text Request