Particle size distributions (PSDs) plays an important role in designing sand control screens. Using different techniques (Dry Sieving, LPSA, and Dynamic Image Analysis (DIA)), large number of PSDs could be measured for core samples in a certain project. Moreover, large-scale sand retention tests are becoming common practice in recent years. These tests usually use duplicated sand mixtures of representative PSDs. Therefore, clustering the PSD data is essential for sand control design and sand retention tests. Supervised and unsupervised machine learning algorithms are getting more attention in computational petroleum engineering. Usually there is no clear idea that how many clusters are supposed to be detected in each PSD database. Therefore, due to the limitation for setting the number of clusters, PSD clustering could not be accomplished using conventional clustering algorithms such as k-means or artificial neural networks. As a new approach, PSD clustering based on an incremental clustering algorithm is used here. The proposed algorithm has online incremental learning capability and it is based on adaptive resonance theory (ART). Besides, the number of clusters is not needed to be assigned as an input parameter in the algorithm. The algorithm, based on a self-adaptation approach, tries to minimize the number of clusters. Accordingly, it is appropriate for PSD clustering of big databases. The proposed algorithm can be used in industrial applications such as sand control design and sand control evaluation testing.
Unsupervised and Self-Adaptive Algorithm for Particle Size Distribution Clustering. Presented at the Geoconvention
Help with Unsupervised and Self-Adaptive Algorithm for Particle Size Distribution Clustering. Presented at the Geoconvention:
