How to successfully migrate a control system

Greg: There are many practical considerations to ensure efficient and productive control system migration. We’re fortunate to talk with Erik Cornelsen, who is the leading contributor to the ISA’s “Ask the Automation Pros” Q & A articles. Erik is an automation and process control engineer at DPS Group, a Scotland-based system integrator. With more than a decade of experience, he’s worked and lived in six countries, contributing to diverse industrial sectors, including food and beverage, logistics and construction materials. He holds a master’s degree in mechanical engineering from INSA de Lyon in France, and is a chartered engineer, member of the Institution of Mechanical Engineers in the U.K., and is an active ISA member.

To begin, what are the ways to migrate code and what must be checked and tested?

Erik: There are two ways to migrate code. The first is done without rewriting the software (auto-migration), and the second involves rewriting it.

After performing auto-migration, verify that all software blocks are identical to the original program. Pay special attention to functions that may have been automatically converted to a different implementation, potentially altering tag names or addresses, and human machine interface (HMI) connections.

If any portion of the code was rewritten, confirm that its behavior matches the original logic. Give extra attention to sequences and non-standard tuning elements (e.g., debounce timers, filters, custom logics). Remove all obsolete or unused input/output (I/O) from the code. Improve the network architecture in new projects where possible (e.g., implement redundant fieldbuses).

Validate that all analog scaling is correctly normalized and scaled. Test as much as possible during the workshop and factory acceptance testing (FAT), including I/O checks, network communications (e.g., fieldbus, couplers, gateways, firewalls), simulation, HMI trends and alarms.

Review proportional-integral-derivative (PID) controller parameters, scan times, type of action (reverse/direct) and control settings. Run a dynamic simulation with proper process relationships and dynamics, and monitor process and PID behavior. Be prepared to fine tune the PID loops.

If other maintenance or cleaning activities are performed during the shutdown, they may also impact the PID settings. For example, if insulation inside a vessel is reduced, the PID controller may need to respond faster and more aggressively, requiring  increased gain.

Ensure setpoint values are retained or saved in the new programmable logic controller (PLC) or distributed control system (DCS). Implement a first-scan routine to initialize them as needed.

Configure correct starting values for all setpoints and parameters.

Review signal filtering logic in the software (e.g., moving averages for analog inputs). Add signal filtering, if necessary, especially since new hardware typically updates and processes data faster than older systems.

Create a comprehensive watch table for all I/O. Pay close attention to signals that operate with inverted logic, such as fail-open valves. Perform a full cross-reference of all I/O to ensure every signal is used correctly, and none were overlooked during implementation.

Confirm the type of each variable-frequency drive (VFD) application to ensure correct configuration and control mode. The most common applications are:

• Constant torque: Torque is constant over the speed range, and power is proportional to speed (screw compressors, screw pumps, conveyors, feeders, extruders, hoists).
• Variable torque: Torque is proportional to the square of the speed, and power is proportional to the cube of the speed (fans, centrifugal pumps). In this type of application, it’s easy to save electricity  by reducing the motor speed slightly.

Add a ramp to smooth and improve some processes, such as large motors reaching target speed, or smoothing the output of a sensitive PID control loop, such as a compressor.

Check PLC and DCS scan times, as well as the central processing unit (CPU) and memory loads. Check the supervisory control and data acquisition’s (SCADA) CPU and memory loads. Configure Windows's auto-login as operator user in the SCADA client’s personal computers (PC).

Greg: How do you  ensure reasonable PID tuning?

Erik: Verify correct PID operation. Consider Greg’s recommended initial tuning parameters (below) from his book, Tuning and Control Loop Performance, Fourth Edition, as a starting point.

When tuning or studying the existing system, perform a step test to identify the deadtime (td), time constant and process gain. It’s also essential to determine the process type (e.g., first-order or second-order self-regulating, integrating process). The deadtime sets the pace for tuning, and methods such as lambda tuning (modified for near-integrating processes) provide a good first approximation. As suggested by Michel Ruel typical starting relationships are:

Scan-rate: tsampling ≈ 0.1 × td

Derivative time: Td ≈ 0.5 × td

Integral time: Ti ≈ 2-4 × td

Also, note that the ISA Standard form PID derivative time must be less than 1/4 integral time, whereas PID Series form inherently enforces this rule.