Real time prediction of drilling fluid rheological properties using Artificial Neural Networks visible mathematical model (white box)

Salaheldin Elkatatny, Zeeshan Tariq, Mohamed Mahmoud

Research output: Contribution to journalArticlepeer-review

142 Scopus citations

Abstract

The drilling fluid rheological properties should be monitored frequently during the drilling operations to prevent the problems related to the change in these properties. Properties such as yield point, plastic viscosity, and apparent viscosity are crucial to evaluate the drilling fluid efficiency in cleaning the well. These properties are only measured twice or once a day, but the Marsh funnel viscosity, solid content, and drilling fluid density are measured every 10 min. Previous models were introduced only to predict the apparent viscosity of the drilling fluid from the Marsh funnel viscosity with large errors. In this paper and for the first time we introduced new model to predict the drilling fluid rheological properties from the Marsh funnel viscosity, solid content, and density measurements in real time. We developed a mathematical model that obtained from the weights, biases, and the transfer functions used in the Artificial Neural Networks (ANNs). The ANNs black box was converted to white box to obtain a visible mathematical model that can be used to predict the drilling fluid rheological properties only using Marsh funnel viscosity, solid content, and density. Based on 9000 data points (collected from the field measurements for actual drilling fluid samples) used in model training and testing, the viscometer readings at 300 and 600 rpm were predicted using the visible mathematical model from the ANNs. The rheological parameters such as yield point, plastic viscosity, apparent viscosity, and consistency index were determined from the viscometer readings at 300 and 600 rpm. The predicted rheological parameters were compared with the measured ones from the field and the match was very good. The average absolute error for the various parameters ranges from 1 to maximum 5 compared to 60 if we used the previously developed correlations. The developed model is a robust technique and tool that can be used to predict the real time drilling fluid rheological parameters that are essential for the drilling hydraulics design and also to predict the performance of drilling fluid. Efficient performance of the drilling fluids depends on the quality of the drilling fluid which needs to be monitored frequently and with the new model this process will be achievable.
Original languageEnglish (US)
Pages (from-to)1202-1210
Number of pages9
JournalJournal of Petroleum Science and Engineering
Volume146
DOIs
StatePublished - Oct 1 2016
Externally publishedYes

Bibliographical note

Generated from Scopus record by KAUST IRTS on 2023-09-20

ASJC Scopus subject areas

  • Geotechnical Engineering and Engineering Geology
  • Fuel Technology

Fingerprint

Dive into the research topics of 'Real time prediction of drilling fluid rheological properties using Artificial Neural Networks visible mathematical model (white box)'. Together they form a unique fingerprint.

Cite this