Secondary Air Injection System Airflow Rate Calculation

Overview of Technical Issues:

The secondary air injection system's airflow rate calculation method is functionally insufficient—it cannot accurately determine the required injection volume needed to optimize catalytic converter performance during cold start and varying engine load conditions, resulting in either incomplete exhaust gas treatment when air delivery is too low or energy waste and potential catalyst thermal damage when air delivery is excessive; the goal is to establish a reliable calculation method that ensures optimal air injection rates across all operating conditions.

Solution directions generated for this problem

Problem Direction 1 :

ImproveAirflow calculation accuracy
VS
ConstraintSystem complexity

Inspiration 1 : Cross-domain reference

Application Principle: #26 Copying
Cross-domain applicability Assess applicability
Fat tree adaptive routing
Innovative Solution Refine solution

Virtual exhaust state observer for model-based airflow calculation

Model-based virtual sensing replaces physical sensors
How to solve :
  • Implement a Luenberger observer that fuses existing engine control signals (throttle position, RPM, ignition timing, fuel injection pulse width) with standard oxygen sensor data to estimate exhaust oxygen concentration and temperature gradients — achieving ±5% accuracy without adding physical sensors
  • Deploy a physics-based exhaust manifold thermal model calibrated offline using engine dynamometer data across 500 operating points (load 0–100%, RPM 800–6000, coolant temp 20–95°C), stored as 12kB lookup tables in ECU flash memory
  • Use recursive least squares adaptation running every 2 seconds to correct model drift based on sparse feedback from existing sensors, maintaining long-term accuracy within ±5% tolerance over 150,000 km vehicle lifetime
Expected Effect : Accuracy ±5%, zero hardware added, ECU memory +12kB only
Risk Control :
  • model calibration coverage insufficient for edge cases
  • sensor drift causes observer divergence over time
  • cold start transient modeling accuracy degradation

Problem Direction 2 :

ImproveAirflow calculation accuracy
VS
ConstraintComputational resource consumption

Inspiration 1 : Cross-domain reference

Application Principle: #1 Segmentation
Cross-domain applicability Assess applicability
Intelligent assistant
Innovative Solution Refine solution

Adaptive multi-resolution airflow calculation with event-triggered precision switching

Divide calculation into coarse and fine modes triggered by operating conditions
How to solve :
  • Implement dual-resolution calculation architecture: coarse mode (±12% accuracy, 10% CPU load) runs continuously at 500ms cycles during steady-state operation
  • fine mode (±5% accuracy, 30% CPU load) activates only when event triggers detect transients—cold start (coolant <60°C), load change (>18% throttle variation within 2s), or catalyst temperature deviation (>25°C from target)
  • Deploy predictive event detector using lightweight decision tree algorithm (5% CPU overhead) monitoring 4 parameters: coolant temperature rate-of-change (≥2°C/s), throttle position derivative, oxygen sensor voltage swing (>0.3V/s), engine RPM acceleration—switching to fine mode 200ms before transient peak, reverting after 3s stability confirmation
  • Store pre-calibrated lookup tables (5D: load 0-100%, RPM 600-6000, coolant 20-95°C, catalyst 150-650°C, ambient -20 to 45°C) generated offline via dynamometer testing covering 10,000 operating points—coarse mode uses 15% grid resolution (800 table entries, 12kB memory), fine mode uses 5% resolution (6,400 entries, 96kB memory) with cubic spline interpolation achieving ±5% accuracy within 80ms
Expected Effect : Average CPU load +18% vs baseline; ±5% accuracy during transients (95% coverage); steady-state load reduced 65%
Risk Control :
  • Event detector false triggers during borderline conditions
  • lookup table interpolation error at grid boundaries
  • memory allocation conflicts with other ECU functions

Problem Direction 3 :

ImproveReal-time measurement precision
VS
ConstraintSystem complexity

Inspiration 1 : Cross-domain reference

Application Principle: #26 Copying
Cross-domain applicability Assess applicability
Fat tree adaptive routing
Innovative Solution Refine solution

Virtual exhaust state observer for model-based airflow calculation

Model-based virtual sensing replaces physical sensors
How to solve :
  • Implement a Luenberger observer that fuses existing engine control signals (throttle position, RPM, ignition timing) with standard O2 sensor data to reconstruct exhaust oxygen concentration and temperature with ±5% accuracy, eliminating need for wideband O2 sensor and fast-response thermocouple
  • Deploy a physics-based exhaust manifold thermal model (1D heat transfer equations with convection coefficient 50–80 W/m²K) calibrated offline using 500-point engine map data, stored in 64KB lookup table for real-time state estimation within 80ms cycle time
  • Use Kalman filter fusion algorithm (5-state vector: exhaust temp, O2 concentration, catalyst bed temp, flow velocity, air-fuel ratio) with measurement update every 200ms from existing sensors and prediction step every 50ms from engine model, achieving effective <50ms response without hardware upgrade
Expected Effect : Accuracy ±5%, sensor count −60%, cost −$45/unit
Risk Control :
  • Model parameter drift over vehicle lifetime
  • calibration map coverage gaps under extreme conditions
  • observer convergence delay during rapid transients

Problem Direction 4 :

ImproveReal-time measurement precision
VS
ConstraintComputational resource consumption

Inspiration 1 : Cross-domain reference

Application Principle: #1 Segmentation
Cross-domain applicability Assess applicability
Intelligent assistant
Innovative Solution Refine solution

Adaptive multi-resolution airflow calculation with event-triggered precision switching

Divide calculation into coarse and fine modes triggered by operating conditions
How to solve :
  • Implement dual-resolution calculation architecture: coarse mode (±10% accuracy, 20% CPU load) runs continuously at 500ms cycles
  • fine mode (±5% accuracy, 60% CPU load) activates only during detected transients—cold start first 45 seconds, throttle change >18%, catalyst temperature gradient >8°C/s—reducing average computational load by 65%
  • Deploy event detection logic using existing ECU sensors (coolant temp, MAP, TPS) with threshold comparisons requiring <2% additional processing, triggering fine mode within one calculation cycle
  • Store pre-calibrated transition lookup tables (4D: load 0-100%, RPM 600-6000, coolant 20-95°C, catalyst 150-600°C) in 128KB flash memory during manufacturing, enabling fine mode to interpolate optimal injection rates in <15ms without iterative algorithms
Expected Effect : ±5% accuracy during transients; average CPU load +18% vs baseline; response <100ms when needed
Risk Control :
  • transition threshold calibration sensitivity
  • lookup table coverage gaps in edge conditions
  • event detection false triggers during noise

Problem Direction 5 :

ImproveCalculation response speed
VS
ConstraintSystem complexity

Inspiration 1 : Cross-domain reference

Application Principle: #10 Preliminary action
Cross-domain applicability Assess applicability
Managing machine to machine devices
Innovative Solution Refine solution

Pre-calibrated multi-dimensional lookup table for rapid airflow injection calculation

Offline pre-compute injection rates across all operating conditions
How to solve :
  • Conduct comprehensive offline calibration testing across 5 dimensions (engine load 0–100%, RPM 600–6000, coolant temp −40–120°C, catalyst temp 200–800°C, ambient temp −30–50°C) with 20-point resolution per axis, generating a 100,000-entry lookup table stored in ECU flash memory
  • Implement 5D linear interpolation algorithm on existing ECU hardware using pre-indexed table structure—calculation completes in <80ms via direct memory access without iterative optimization
  • Validate table accuracy through dynamometer testing across 50 drive cycles, ensuring ±5% injection precision with acceptance criteria requiring 95% of test points within tolerance
Expected Effect : Calculation speed <80ms; no hardware upgrade; ±5% accuracy maintained
Risk Control :
  • calibration coverage gaps in edge conditions
  • flash memory wear from table updates
  • interpolation error accumulation

Problem Direction 6 :

ImproveCalculation response speed
VS
ConstraintComputational resource consumption

Inspiration 1 : Cross-domain reference

Application Principle: #10 Preliminary action
Cross-domain applicability Assess applicability
Fact checking method and system utilizing format
Innovative Solution Refine solution

Offline pre-calibrated 5D lookup table for rapid airflow calculation

Pre-calibrate airflow injection maps offline to shift computational burden from runtime
How to solve :
  • Generate comprehensive 5D lookup tables (engine load 0-100%, RPM 600-6000, coolant temp -40°C to 120°C, catalyst temp 200-900°C, ambient temp -30°C to 50°C) during factory calibration using high-performance workstations, covering 95% operating conditions with 10% grid resolution
  • Store tables in ECU flash memory (2MB allocation) with indexed addressing structure enabling trilinear interpolation retrieval in <15ms per query
  • Implement lightweight boundary detection algorithm (5ms overhead) to identify edge cases requiring fallback calculation, maintaining ±5% accuracy across normal operation while reducing real-time processing load by 85%
Expected Effect : Calculation cycle <100ms; ECU load -85%; accuracy ±5%
Risk Control :
  • lookup table memory overflow risk
  • interpolation error at grid boundaries
  • calibration coverage gaps in extreme conditions
Patsnap Eureka Solution