Background:
Munters, a global leader in indoor climate control, is striving to always provide customers with the most precise and energy efficient equipment possible. The purpose of a climate control system is to maintain a certain humidity, temperature, or air quality required for e.g., production or storage of sensitive products. The control system should reliably attenuate disturbances and quickly track setpoint changes, to maintain the desired, stable indoor climate.
Over time, site conditions and seasonal loads can cause control loops to drift from optimal tuning. Poorly tuned controllers increase energy use, reduce comfort and reliability, and raise service costs. One of the most critical issues with poorly tuned controllers is that they can prevent us from meeting the customer’s climate requirements such that they can’t reliably produce their goods due to unstable conditions. The goal is to create a continuous, data-driven performance assessment framework that can automatically flag when a loop requires re-tuning.
Research Questions:
1. How can closed-loop performance be quantified continuously using only operational data?
2. Which performance indices are most reliable for detecting suboptimal tuning in humidity control loops?
3. How can disturbance effects (e.g., outdoor humidity changes, site condition changes) be separated from tuning issues?
Project Objectives:
Disturbance handling for detecting “calm” windows:
Investigate and compare methods for distinguishing calm (low-disturbance) operating periods from transient or load-driven conditions.
Potential approaches include:
Ambient-rate gating: exclude periods with large environmental changes (e.g., variations in process inlet or outdoor humidity).
Model-based residual analysis: train a simple predictive model between key inputs (e.g., heater signal, process air humidity) and the controlled humidity, then use residual magnitude as a calmness indicator.
The student is encouraged to evaluate these and possibly propose or combine alternative criteria that improve robustness or automation.
undefined
Performance indices (evaluated on calm windows):
Explore suitable performance metrics for assessing control quality and detecting sub-optimal tuning.
Candidate indices include, but are not limited to:
Harris (minimum variance) index: benchmark variance ratio.
Oscillation index: spectral peak strength of control error.
Control effort index: RMS/variation of actuator activity relative to control error.
Steady-state bias: median offset error.
The student may introduce or adapt other indices found in literature or discovered through data analysis.
undefined
Retuning rule:
Develop and test rules for flagging when re-tuning might be required, based on one or more of the selected indices.
Evaluate sensitivity and robustness across sites and operating conditions.
Expected Outcomes:
Identification and evaluation of suitable methods for distinguishing calm (low-disturbance) operating periods.
Selection and assessment of relevant performance indices.
Development of a data-driven framework that integrates disturbance handling and control performance assessment.
Demonstration and validation of the framework on at least one real data set, including example cases illustrating when re-tuning would be recommended.
Qualifications: MSc student in Control, Automation, or Applied Physics/Engineering with background in:
Control system principles
System identification and/or time-series modeling
Signal processing for performance monitoring
Programming skills (Python preferred)
Contact: Shervin Parvini Ahmadi, Research Engineer
Email: shervin.parvini.ahmadi@munters.com