No abstract

Accurate estimation of flow in drilling operations at inflow and outflow positions can lead to increased safety, optimized production and improved cost efficiency. In this paper, Dynamic Artificial Neural Network (DANN) is used to estimate the flow rate of non-Newtonian drilling fluids in an open channel Venturi-rig that may be used for estimating outflow. Flow in the Venturi-rig is estimated using ultrasonic level measurements. Simulation study looks into fully connected Recurrent Neural Network (RNN) with three different learning algorithms: Back Propagation Through Time (BPTT), Real-Time Recurrent Learning (RTRL) and Extended Kalman Filter (EKF). The simulation results show that BPTT and EKF algorithms converge very quickly as compared to RTRL. However, RTRL gives more accurate results, is less complex and computationally fastest among these three algorithms. Hence, in the experimental study RTRL is chosen as the learning algorithm for implementing Dynamic Artificial Neural Network (DANN). DANN with RTRL learning algorithm is compared with Support Vector Regression (SVR) and static ANN models to assess their performance in estimating flow rates. The comparisons show that the proposed DANN is the most accurate model among three models as it uses previous inputs and outputs for the estimation of current output.

In oil and gas and geothermal installations, open channels followed by sieves for removal of drill cuttings, are used to monitor the quality and quantity of the drilling fluids. Drilling fluid flow rate is difficult to measure due to the varying flow conditions (e.g., wavy, turbulent and irregular) and the presence of drilling cuttings and gas bubbles. Inclusion of a Venturi section in the open channel and an array of ultrasonic level sensors above it at locations in the vicinity of and above the Venturi constriction gives the varying levels of the drilling fluid in the channel. The time series of the levels from this array of ultrasonic level sensors are used to estimate the drilling fluid flow rate, which is compared with Coriolis meter measurements. Fuzzy logic, neural networks and support vector regression algorithms applied to the data from temporal and spatial ultrasonic level measurements of the drilling fluid in the open channel give estimates of its flow rate with sufficient reliability, repeatability and uncertainty, providing a novel soft sensing of an important process variable. Simulations, cross-validations and experimental results show that feedforward neural networks with the Bayesian regularization learning algorithm provide the best flow rate estimates. Finally, the benefits of using this soft sensing technique combined with Venturi constriction in open channels are discussed.

This paper presents a flexible system structure to analyze and model for the potential use of huge ship sensor data to generate efficient ship motion prediction model. The noisy raw data is cleaned using noise reduction, resampling and data continuity techniques. For modeling, a flexible Support Vector Regression (SVR) is proposed to solve regression problem. In the data set, sensitivity analysis is performed to find the strength of input attributes for prediction target. The highly significant attributes are considered for input feature which are mapped into higher dimensional feature using non-linear function, thus SVR model for ship motion prediction is achieved. The prediction results for trajectory and pitch show that the proposed system structure is efficient for the prediction of different ship motion attributes.

Open channel flow of complex fluids is found in many offshore applications and is currently monitored using Coriolis meters (good uncertainty with an expensive device) and simple paddle meters (very poor uncertainty). Recent publications in IEEE by the current authors indicate that the flow of complex fluids in open channels can be estimated by level measurements in the open channel by scanning the surface of the fluids in the open channel with an array of ultrasonic sensors. Complex fluids possess rheological properties dependent on flow, density, pipe dimensions etc. As an interesting industrial application of different types of sensors, this paper presents the basic configuration of the sensors used in a pilot scale study with some selected samples of complex fluids. A comparison of the performances of the sensors using coefficient of variations (CV) with respect to the mean values of the measurands is given as a preamble before using them in the final mass flow estimation. The group on multiphase studies in USN recently used various statistical parameters in the identification of flow regimes in multiphase flow studies as reported in the IEEE Sensors Community. In addition, the measurand values are filtered using different algorithms. The flow in the open channel is estimated using a Radial Basis Neural Network (RBNN) with the levels from the ultrasonic scanning array as inputs and the mass flow as output. The paper summarizes the findings with some indications of their implications to the offshore and other industries.