Control Performance Monitoring (CPM)

Compares current loop performance against the theoretical minimum variance achievable given the process's inherent time delay. The Harris Index quantifies this ratio, providing a universal metric between 0 and 1. A score near 1 indicates near-optimal control, while a declining index signals increasing loop sluggishness or excessive actuator movement. This benchmark requires only routine operating data and knowledge of the process dead time, making it non-intrusive.

Automatically identifies persistent cyclical variability in process data using techniques like autocorrelation functions and power spectral density analysis. CPM distinguishes between common oscillation sources:

• Tuning-induced: Aggressive controller gains causing limit cycles
• External disturbances: Upstream process variability propagating downstream
• Stiction: Valve static friction causing stick-slip cycles
• Interaction: Coupling between improperly decoupled multi-variable loops

Detects the characteristic signature of static friction in pneumatic control valves by analyzing the OP-PV plot (controller output vs. process variable). Stiction manifests as a distinct parallelogram or elliptical pattern. CPM quantifies the stiction band—the minimum change in controller output required to overcome static friction—and estimates the resulting economic loss from increased variability and actuator wear.

Uses routine closed-loop data to identify a process model and estimate key performance indicators:

• Settling time: Time to recover from a disturbance
• Overshoot: Maximum deviation beyond setpoint
• Output variance: Standard deviation of the controlled variable
• Actuator travel: Cumulative valve movement indicating wear These metrics are trended over time to detect gradual performance degradation before alarms trigger.

Constructs a cross-correlation matrix across all monitored loops to identify hidden interactions and propagation paths. A disturbance in one loop may manifest as variability in another, leading operators to misdiagnose the root cause. CPM's interaction mapping reveals the true source loop, enabling targeted corrective action rather than wasting effort on the symptomatic loop.

Aggregates performance indices into a dashboard hierarchy that ranks loops by economic impact. Poorly performing loops on critical streams—such as reactor temperature or distillation column reflux—are flagged with higher severity. Reports include:

• Pareto charts of worst-performing loops
• Trend plots with statistical control limits
• Prescriptive recommendations for tuning or maintenance This transforms raw diagnostics into actionable work orders for instrument technicians and process engineers.

The theoretical gold standard for CPM. This metric calculates the lowest achievable output variance if a controller were perfectly designed, using only the process time delay. The Harris Index compares actual variance to this minimum, providing a normalized performance ratio between 0 and 1. A value near 1 indicates optimal control, while lower values signal degradation from valve stiction, sensor drift, or poor tuning.

A core CPM diagnostic that automatically identifies persistent cyclic variability in process data. Algorithms like the autocorrelation function (ACF) and power spectral density analysis distinguish between external oscillatory disturbances and internally generated limit cycles. Key indicators include:

• Decay ratio: Quantifies damping of oscillations
• Periodicity index: Measures regularity of the cycle
• Valve stiction patterns: Saw-tooth waveforms indicating mechanical friction

The most common mechanical root cause of poor loop performance. CPM systems detect stiction by identifying stick-slip patterns in the controller output (OP) and process variable (PV) relationship. The characteristic signature is a parallelogram-shaped phase plot where the valve stem sticks before abruptly slipping. Advanced diagnostics quantify the stiction band (percentage of input span) and distinguish it from external disturbances or aggressive tuning.

A complementary discipline to CPM that monitors product quality characteristics rather than loop dynamics. While CPM evaluates controller health, SPC uses control charts (X-bar, R, EWMA) to distinguish common cause variation (inherent to the process) from special cause variation (assignable to a specific event). Together, they provide a complete view: CPM ensures the loop is healthy; SPC ensures the output meets specifications.

Extends traditional CPM to multi-variable controllers. MPC-specific diagnostics track:

• Model mismatch: Divergence between predicted and actual plant behavior
• Constraint activity: Frequency and duration of operating at limits
• Ill-conditioning: Sensitivity of the optimization to small data changes
• Rank deficiency: Loss of independent manipulated variables These metrics detect when an MPC requires re-identification or redesign.

The optimal recursive algorithm underlying many CPM diagnostics. A Kalman filter estimates the true internal state of a process from noisy measurements by minimizing the mean squared error. In CPM, it enables:

• Innovation sequence analysis: White noise residuals indicate a well-tuned filter
• Covariance monitoring: Detects changes in process noise characteristics
• Sensor fault detection: Identifies bias, drift, or precision degradation in individual instruments