CNC Mill Concept: Compliance, State Estimation & Supervision

Introduction

This post is part of the CNC Mill Concept hub. It covers quasi-static compliance estimation, ECU-based machine state classification, the supervision logic that acts on those states, and the state transition hysteresis model.

Compliance Estimation and Structural Observability

Purpose

• Estimate quasi-static structural behavior under load
• Detect excessive compliance that risks tolerance loss
• Provide slow-varying state information for process supervision
• Not intended for real-time control loops

Tilt Estimation

• Derived from accelerometer data
• Uses gravity vector for orientation reference
• Strong low-pass filtering applied
• Valid only in quasi-static conditions
• Used to estimate spindle plate rotation under load

Strain Gauge Measurement

• Strain gauge mounted on X carriage
• Measures bending of X gantry under load
• Acts as cutting-force proxy
• Provides information independent of vibration measurements

Motor Current as Load Proxy

• Motor current monitored at spindle drive
• Two components used:
  • fast component:
    • incipient stall detection
  • slow envelope:
    • mean cutting load estimation

Compliance Model

• Structural compliance model maintained in ECU
• Model inputs:
  • spindle plate tilt
  • X-carriage strain
  • motor current
• Model output:
  • estimated tool/spindle deflection envelope
• Model characteristics:
  • calibrated using known static or quasi-static loads
  • evaluated only in steady-state conditions
  • versioned and recalibratable
• Model is:
  • a slow observer
  • not part of safety system

Machine State Estimation and Process Supervision

Purpose

• Combine sensor data into a coherent machine state
• Supervise the machining process based on estimated state
• Prevent tolerance loss, chatter, and spindle stall
• Never perform motion planning or trajectory generation

Machine State Estimation

• State estimation performed in ECU
• Inputs:
  • vibration:
    • piezo surface microphones
    • accelerometers
  • quasi-static tilt:
    • low-pass-filtered accelerometer data
  • structural load:
    • X-carriage strain gauge
  • spindle load:
    • motor current
• Sensor fusion performed with explicit time-scale separation:
  • fast signals:
    • vibration
    • chatter indicators
  • slow signals:
    • tilt
    • strain
    • motor current envelope

Machine State Classification

• ECU classifies the current operating state as:
  • normal operation
  • static overload / excessive compliance
  • dynamic instability (chatter)
  • incipient spindle stall

Process Supervision Actions

• ECU supervises process only
• ECU does not:
  • control axis motion
  • modify trajectories
  • override safety systems
• Actions requested via:
  • Duet macros
  • GPIO signals for low latency

Supervision Logic

• Static overload / excessive compliance:
  • detected by increased tilt and/or strain
  • action: reduce feed rate
• Dynamic instability (chatter):
  • detected by elevated vibration energy
  • action: adaptive spindle speed adjustment
• Incipient stall:
  • detected by rapid motor current increase
  • action: immediate feed reduction

State Transitions and Hysteresis

• State transitions implemented using Schmitt-trigger logic
• Upper threshold:
  • spindle motor current ≥ 100 % rated
  • transition to reduced-feed state
• Lower threshold:
  • spindle motor current ≤ configurable value (e.g. 80 %)
  • maintained for configurable time
  • transition back to normal-feed state
• Hysteresis prevents oscillation and hunting

Configuration and Scope

• All thresholds and timing parameters:
  • configurable via Raspberry Pi plugin
  • logged for analysis
• Supervisory logic:
  • non-safety-critical
  • fully disable-able
  • degrades gracefully if unavailable
• Safety functions remain hardware-only and independent

Architecture Diagram

Fig. 1 shows the full state estimation and supervision architecture with sensor inputs, ECU processing layers, and output connections to the Duet and spindle drive.

License

This project (with exceptions) is published under the CC Attribution-ShareAlike 4.0 International License.