Construction and Application of Digital Twin Model of Oil Rig Lifting System in Industrial Internet of Things

As the core equipment in the oil drilling process, oil drilling rigs play a vital role in drilling operations. Digital technology upgrades for oil drilling rigs can achieve the goal of improving the automation level of drilling operations. By tapping the potential of Internet of Things (IoT) systems, using the data they provide, and using digital twin technology to digitally upgrade oil drilling rigs, drilling efficiency can be effectively improved and production losses can be reduced. Digital twins refer to the creation of digital copies of the real world or entities through digital technology, with the goal of simulating, monitoring, and optimizing the performance of physical entities. This technology can create virtual models by collecting multi-source data, analyzing the stress of objects under different stress conditions, and deducing the deformation of objects. At present, the accuracy of digital twin modeling is affected by the different sources of sensor data, different data formats, and inconsistent data accuracy. In response to the multi-source and heterogeneous problems of sensor data, we have built a multi-source heterogeneous data fusion model based on multiple sensors based on the multi-source heterogeneous data fusion model in physical space. To support this model, we proposed a convolutional neural network PSO-CNN-SA that integrates a self-attention mechanism. In this network, we introduced a particle swarm optimization algorithm to quickly initialize network parameters, and combined it with the self-attention mechanism to adjust the weights between heterogeneous data, thereby improving the positioning accuracy of data fusion. At present, due to the different sources of sensor data, different data formats, and inconsistent data accuracy, the PSO-CNN-SA network uses the carbon fiber arm model of the oil rig as the experimental object. After simulation fusion, the positioning accuracy is improved by 72.6%. The experimental results confirm the feasibility of the network architecture in multi-source heterogeneous data fusion tasks and also provide reliable data support for subsequent digital twin dynamic deduction.