Experimental Study of Shape Memory Sand Control Screen

Abstract

Shape memory sand control screen completion based on shape memory polymer not only has the advantages of simple process and easy to run in hole, like independent screen, but also can achieve the sand management effect of gravel filling. Therefore, shape memory sand control screen has wide application prospects. However, since the shape memory material is temperature-sensitive, a large number of laboratory experiments are needed to evaluate its expansion, seepage and sand retaining capabilities, as well as optimize the polymer system and screen structure. A performance evaluation experiment system for full-size shape memory screen prototype was developed. The device can describe the parameters of the shape memory screen during the expansion process and after expansion in real time and quantitatively, such as permeability, outer diameter and residual stress. The expansion behavior of screen prototype is controlled by step heating the passing through fluid to simulate the screen run in hole operation. 60 sets of experiments were carried out using the device. The expansion performance, seepage performance and sand control performance of shape memory screens were evaluated. Shape memory polymer formulation and screen structure are also optimized. The research shows that the optimized shape memory sand control screen densely filled annulus and effectively supports the wellbore after expansion. The permeability of the expanded screen can be up to 35µm2, the displacement pressure difference can be less than 1kPa, and the sand control precision can be up to 0.061mm. Therefore, shape memory sand control screen can be used for sand control completion of oil and gas wells with limited well site conditions such as long horizontal wells, complex wells and offshore wells. This paper presents an experiment device to evaluate the performance of shape memory screen, the dynamic expansion behavior is described during the expansion process and after expansion in real time and quantitatively.

Application of a CNN to the Boda Claystone Formation for high-level radioactive waste disposal

“…The hyperparameters of the network (e.g., learning rate) are subtle but potentially vital details in the ultimate learning process 63 . ArcGIS Pro's "Train Deep Learning Model" geoprocessing tool does not offer optimizer selection on its user interface, but this function was identified in the underlying packages 54,58 that the training tool calls an exact open-source FastAI code, which has adaptive moment estimation (' Adam'; 64 ) as its default 65,66 . Adam is one of the so-called "adaptive optimizers", achieving finer networks by specializing a particular learning rate for each parameter 64 .…”