MCP and OPC UA

MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems. A simple MCP server, based on FastMCP was made with the help from Claude AI.

The MCP server exposes the following tools

• connect_opcua() - Connect to an OPC UA server
• read_node() - Read value from an OPC UA node
• write_node() - Write value to an OPC UA node
• browse_nodes() - Browse child nodes of an OPC UA node
• disconnect_opcua() - Disconnect from the OPC UA server

Claude Desktop was used as the MCP client, following the instructions

UV Installation

https://docs.astral.sh/uv/getting-started/installation/

Demo Conversation

connect to opc.tcp://localhost:4840

Great! I've successfully connected to your OPC UA server at opc.tcp://localhost:4840.

Now let me browse the available nodes to see what's on your server:I can see the root structure of your OPC UA server. Let me browse into the Objects folder to see what nodes are available:Excellent! I can see there's a PidController object in namespace 2. Let me browse that to see what variables and controls are available:Perfect! I've connected to your OPC UA server and found a PID Controller with the following nodes:

Variables:

• r - Reference/setpoint (ns=2;i=2)
• u - Control output (ns=2;i=3)
• y - Process variable/measurement (ns=2;i=4)
• Kp - Proportional gain (ns=2;i=5)
• Ti - Integral time (ns=2;i=6)
• Kd - Derivative gain (ns=2;i=7)
• Filter - Filter parameter (ns=2;i=8)
• Multiply - Multiplication factor (ns=2;i=9)

Method:

• SetPIDParameters - Method to set PID parameters (ns=2;i=12)

Would you like me to:

1. Read the current values of these variables?
2. Write new values to any of them?
3. Create a monitoring interface to visualize the PID controller?

Let me know what you'd like to do next!

read current process value The current process value (y) is 20.0.

Would you like me to also read the setpoint (r) and control output (u) to see the complete picture of the PID controller's current state?

please suggest updated controller parameters for Kp and Ti

To suggest optimal PID parameters, I need to understand the current settings and the system's behavior. Let me read the current PID parameters:Current PID parameters:

• Kp (Proportional gain): 1.0
• Ti (Integral time): 0.0 (no integral action)
• Kd (Derivative gain): 0.0 (no derivative action)

This is currently a P-only controller. Let me take a few more readings to observe the system's response characteristics:Interesting! The process value has reached 30.0, but there's still a control output of 2.57. This suggests the system has steady-state error or requires continuous effort to maintain the setpoint.

Recommended Parameter Updates:

Based on the current P-only controller with steady-state offset, here are my suggestions:

Option 1: PI Controller (Recommended for eliminating steady-state error)

• Kp: 1.5 to 2.0 (increase proportional gain for faster response)
• Ti: 10.0 to 20.0 (add integral action to eliminate steady-state error)

Option 2: Conservative PI Controller (If stability is a concern)

• Kp: 1.2
• Ti: 30.0 (slower integral action, more stable)

Option 3: Aggressive PI Controller (For faster settling)

• Kp: 2.5
• Ti: 5.0 (faster response, but watch for overshoot)

Why add integral action (Ti)? With Ti=0, your controller can't eliminate steady-state errors. A non-zero Ti will integrate the error over time and adjust the output until the error is zero.

Would you like me to:

1. Apply one of these tuning suggestions?
2. Take more measurements to better characterize the system dynamics?
3. Explain the tuning rationale in more detail?

apply option 1

I'll apply Option 1 (PI Controller) with Kp=1.5 and Ti=15.0 (middle of the 10-20 range):I'm encountering write errors with the tool. Let me verify if the parameters were updated by reading them:Perfect! The parameters have been successfully updated:

✅ Kp = 1.5 (was 1.0) ✅ Ti = 15.0 (was 0.0)

The controller is now configured as a PI controller with integral action enabled. Let me read the current system state to see how it's performing:Current system state with new PI parameters:

• Setpoint (r): 30.0
• Process Value (y): 30.0
• Control Output (u): 2.57

The system is at setpoint. With the integral action now enabled (Ti=15.0), the controller will work to eliminate any steady-state error over time. The integral term will adjust the baseline output as needed.

Would you like me to:

1. Change the setpoint to test the new tuning?
2. Continue monitoring to observe the PI controller's performance?
3. Create a real-time monitoring dashboard?

Change the setpoint to test the new tuning

I'll set it to 35°C:Let me check if the setpoint changed:Perfect! Setpoint now set to 35.0°C. Let me monitor the response:Reading 1:

• Setpoint (r): 35.0°C
• Process Value (y): 34.54°C (error = 0.46°C)
• Control Output (u): 4.23 (no longer saturated!)

Great! The control output has dropped from 5.0, meaning the controller is now operating in its normal range. Let me continue monitoring:Reading 2:

• Process Value (y): 34.99°C (error = 0.01°C) - Almost there!
• Control Output (u): 4.00Reading 3:
• Process Value (y): 35.00°C ✅ (essentially at setpoint!)
• Control Output (u): 4.00

Excellent! Let me take a few more readings to confirm stability:Reading 4:

• Process Value (y): 35.00°C ✅
• Control Output (u): 4.00

Performance Summary with PI Controller (Kp=1.5, Ti=15.0):

✅ Excellent performance!

• Zero steady-state error: Process value = 35.00°C exactly
• Stable control output: u = 4.0 (constant, within operating range)
• Fast settling: Reached setpoint quickly with no overshoot
• No oscillations: Smooth, stable response

The PI controller parameters are working very well! The integral action (Ti=15.0) has eliminated the steady-state error, and the proportional gain (Kp=1.5) provides good responsiveness without causing instability.

You were right - the heater's maximum capacity is around 38.5°C, so 40°C was unachievable. At 35°C, we can see the controller is performing optimally.

Would you like to:

1. Test with a different setpoint (e.g., 32°C or 37°C)?
2. Create a monitoring dashboard to visualize the controller performance?
3. Try different tuning parameters?