Abstract: Based on the nonlinear characteristics of the automatic drill pipe discharge system and the need to ensure the smoothness of the discharge velocity, a fuzzy neural network PID controller suitable for process identification is constructed to identify the object model of the manipulator, providing a nonlinear relationship model for dynamic programming. A real-time controller combining a fuzzy neural network controller and a robust controller is employed to ensure the stability of the closed-loop system and good control performance. Simulation results show that the improved fuzzy neural network PID controller has fast dynamic response, small overshoot, good robustness, and high accuracy.
Keywords: drilling; fuzzy control; PID control; robust control
Intermediate Classification Number: TP9 Document Identification Code: B
ApplicationResearchofOffshoreDrillingBasedonFuzzyNeuralNetwork
WANGShaoYuan
(QingdaoInstituteofRubberTires,Qingdao266042,China)
Foreword
With increasing demand for oil and gas resources, countries have gradually shifted their focus to the deep sea. As operating water depth increases, the number of drill pipes required for drilling also increases, leading to higher drilling time and costs. Traditional drill pipe operation methods are no longer sufficient to address these issues, necessitating the adoption of mechanized and automated drill pipe operating systems. Currently, automatic drill pipe discharge systems are widely used in deep-sea drilling rigs, achieving significant success in improving the safety and economy of drilling operations.
During drilling, drill pipe needs to be transferred from the platform deck to the second-level platform, and individual drill pipes are continuously joined together to form a stand and then attached to the drill string to keep drilling ongoing. To change drill bits and install testing or other equipment at the bottom of the drill string, the drill string needs to be frequently pulled out of the wellbore. After installing a new drill bit or other equipment, the drill pipe needs to be lowered back into the wellbore. This process involves frequent drill pipe retrieval operations, making it a highly repetitive and labor-intensive process.
In the 1940s, researchers began exploring mechanized methods for tripping in and out of the well, leading to the development of automatic drill pipe discharge systems. Currently, these systems are widely used in deepwater and ultra-deepwater offshore drilling rigs, achieving significant success in improving the safety and economy of drilling operations. The tripping in and out of the well requires moving the drill pipe from the wellhead to the drill pipe discharge rack or vice versa – this is the drill pipe discharge operation. This operation is inherently dangerous, highly repetitive, and labor-intensive, requiring substantial time and the cooperation of multiple personnel. Due to the frequent winds and waves encountered in offshore drilling operations, the working environment is harsh, making safety even more critical. Furthermore, offshore drilling is very costly. Therefore, accelerating drilling speed and shortening the drilling cycle are of paramount importance.
1. Introduction to the Automatic Drill Pipe Discharge System
The actual system of a marine platform manipulator exhibits significant nonlinearity, making it highly sensitive to parameter changes and difficult to achieve accurate positioning. Furthermore, due to limitations in mechanical precision, a dead zone exists, reducing control accuracy and causing overshoot. A fuzzy neural network PID controller utilizes a dynamic recurrent neural network (ELM network) to identify the system model, combining fuzzy control with a neural network. The neural network implements fuzzy logic while leveraging its self-learning capability to dynamically adjust membership functions, optimize control rules online, and adjust PID control parameters online. This allows the controller to possess both the self-learning capabilities of the fuzzy neural network and the full advantages of PID control. When the NNI has deviations or the learning is not yet fully converged, a robust feedback controller ensures the stability of the closed-loop system during the initial learning phase of the fuzzy neural network.
The automatic drill pipe discharge control system is mainly divided into a logic control section and a closed-loop control section. Its operation consists of five parts: drill pipe deck operation, drill pipe conveying operation, drill pipe loading and unloading on the drill rig, drill pipe discharge on the drill rig, and single drill pipe connection and erection operation. Its main working principle is that the host computer inputs a position signal to the electro-hydraulic servo system, which then controls the position positioning, gripping, and lowering of the manipulator, with sensors transmitting the manipulator's position signal in real time. The working principle diagram of its closed-loop control section is shown in Figure 1.
Figure 1. Closed-loop control principle diagram of the automatic drill pipe discharge system
2. Improved Fuzzy Neural Network PID Control
2.1 Control System Structure
The control system structure, which is constructed in conjunction with the drill pipe automatic discharge system, is shown in Figure 2. The position of the robot arm is used as the controlled variable.
Figure 2. Structure diagram of the improved fuzzy neural PID control
In the figure, e and ec represent the error and the rate of change of error, respectively. The input r is the position of the robot arm, and the output y is the actual output of the robot arm.
2.2 Structure of Fuzzy Neural Network
The fuzzy neural network has four layers, as shown in Figure 3. Layer 1 is the input layer; layer 2 is the fuzzification layer; layer 3 is the fuzzy inference layer; and layer 4 is the output layer. The fuzzy neural network structure is 2–6–6–3.
Figure 3. Structure of a fuzzy RBF neural network
(l) Input Layer. This layer takes the input error e and the actual system output y(k) as inputs to the next layer. The activation function is:
Therefore, the outputs of this layer are e and y(k).
The fuzzy layer. The activation function is the membership function. Therefore, the output is:
Where i = 1, 2; j = 1, 2, ..., 6. cij and bij are the mean and standard deviation of the membership functions of the j-th fuzzy set of the i-th input variable of the Gaussian function, respectively.
(3) Fuzzy Inference Layer. The output value of this layer is obtained by multiplying the fuzzy quantities in the upper layer pairwise. Therefore, the activation function of this layer, i.e., the output, is:
Here, k = 1, 2, 3, 4, 5, 6.
Output layer. This layer outputs the parameters of the PID controller. The output value of this layer is the weights multiplied by the output of the third layer using a matrix multiplication method. Therefore, the output of this layer is:
The control quantity of incremental PID control is
The objective function is:
Where r(k) is the desired output.
2.3 Identification of Dynamic Recursive Network Models
The nonlinear dynamics of the system are described using the internal state feedback of the network, and the model of the automatic drill pipe discharge system is identified online. A three-layer recurrent neural network is shown in Figure 4.
Figure 4 Recurrent Neural Network Structure
The hidden layer nodes of the network have dynamic recursive connections, and their input-output relationships are as follows:
Input layer:
Hidden layer:
Output layer:
In the formula: Xi(k) — input node of the network; Sj(k), Hj(k) — input and output of the hidden layer; y^(k) — network output; f(x) — activation function, f(x)=1(1+ex); w1ij,w2j,w3j — connection weight vectors between the input layer and the hidden layer, and recursively to the output; k=1,…,m, where m is the number of learning patterns. The function for adjusting the network weights is defined as:
In the formula: y(k) — actual system output; y^(k) — network output.
The adjustment formula for the weights of a dynamic recursive network can be obtained using the gradient descent method:
Therefore, the gradient information of the controlled object is:
In the above formula:
Equations (1) and (2) are nonlinear dynamic recursive equations that vary with time, and can be obtained using known initial conditions:
Obtained recursively.
2.3 Robust Controller
To ensure the stability and good control performance of the closed-loop system, the real-time controller consists of a fuzzy neural network controller (NNC) and a robust controller (RC). The output signals of these two controllers are weighted and synthesized, and then used as the control input of the system to form a variable robust controller u(k):
In the formula: un(k) is the output of the NNC; ur(k) is the output of the robust controller; γ is the identification accuracy of the system model NNI, called the robustness factor. γ is expressed as:
In the formula: τ is the variable robustness coefficient of the robustness factor; Em is the square of the difference between the NNI output and the actual system output.
2.4 PID Control Algorithm
The controller algorithm is as follows:
Traditional incremental PID control algorithm:
3 Simulation Study of Automatic Drill Pipe Discharge System
To verify the effectiveness of the proposed fuzzy neural network PID control algorithm, a fuzzy neural network was created in MATLAB. The abstract fuzzy rules were transformed into training samples for the fuzzy neural network using membership functions and fuzzy rules. The hidden layer used the Tansig function, which is differentiable at any point, as the transfer function, and the output layer used the commonly used non-negative Sigmoid function.
Figure 5 System step response curve
Figure 5 shows the response curves of the system to a step signal when using conventional PID control and improved fuzzy neural network PID control. As can be seen from Figure 5, the improved fuzzy neural network PID control effectively suppresses system overshoot, has a fast response, and small steady-state error, demonstrating performance far superior to conventional PID control. Figure 6 shows the error response curves of the conventional PID controller and the improved fuzzy neural network PID controller for sinusoidal signal tracking. The comparison shows that the fuzzy neural network PID controller significantly outperforms the conventional PID controller in dynamic performance, reducing the sinusoidal response error from 0.02 rad to 0.001 rad.
(a) PID control (b) Neural network PID control
Figure 6. System sinusoidal error response curve
4. Conclusion
The improved fuzzy neural network PID controller combined with a robust controller is presented in this paper. Experimental results using a MATLAB simulator show that the system can shorten the settling time, accelerate the response speed, improve stability accuracy, and has good robustness. It can meet the requirements of nonlinear systems and has practical application value.
