How to Detect Leaks in Absorption Refrigerator Systems

Overview of Technical Issues:

The detection equipment inadequately measures and identifies refrigerant leaks in the absorption refrigerator system, causing small leaks to remain unnoticed until significant refrigerant loss occurs, resulting in degraded cooling performance and potential complete system failure before operators can intervene with repairs.

Solution directions generated for this problem

Problem Direction 1 :

ImproveDetection sensitivity threshold
VS
ConstraintEquipment system complexity

Inspiration 1 : Cross-domain reference

Application Principle: #26 Copying
Cross-domain applicability Assess applicability
Systems and methods to detect rare mutations and copy number variation
Innovative Solution Refine solution

Thermal signature mapping for refrigerant leak detection via infrared imaging

Single infrared camera replaces multi-sensor arrays
How to solve :
  • Deploy uncooled microbolometer infrared camera (8-14μm spectral range, ≤50mK thermal sensitivity) to scan entire refrigerator system surface every 5 minutes, detecting 0.2-0.5°C temperature anomalies caused by refrigerant evaporative cooling at leak sites as small as 5g/year
  • Establish baseline thermal profile during commissioning by capturing 20-30 thermal images over 48 hours under normal operation, then apply pixel-by-pixel differential analysis (threshold: ΔT ≥0.3°C persisting >15 minutes) to identify leak-induced cold spots while filtering transient temperature fluctuations from defrost cycles or ambient changes
  • Mount camera at fixed position 2-3 meters from system with automated scanning protocol—capture full-system thermal map in 60-90 seconds, process images using edge-detection algorithms to highlight anomaly zones, and trigger visual/audible alarms when cold spots match leak signature criteria (size 10-50mm², temperature depression 0.3-1.0°C, persistence >3 consecutive scans)
Expected Effect : Sensitivity 5-10g/year; single device replaces 5-10 sensors; false alarm rate <5%; detection cycle 5-15 minutes
Risk Control :
  • ambient temperature interference affecting baseline accuracy
  • camera calibration drift over 6-12 months
  • reflective surfaces causing thermal imaging artifacts

Problem Direction 2 :

ImproveDetection sensitivity threshold
VS
ConstraintOperational false alarm rate

Inspiration 1 : Cross-domain reference

Application Principle: #23 Feedback
Cross-domain applicability Assess applicability
System and method for presence detection in an environment to be monitored
Innovative Solution Refine solution

Adaptive baseline learning system for refrigerant leak detection with dynamic threshold adjustment

Self-learning baseline system for leak detection
How to solve :
  • Deploy machine learning baseline algorithm that monitors each refrigerator's pressure and temperature patterns continuously for 2-4 weeks commissioning period, establishing unit-specific normal behavior signatures including defrost cycles, compressor starts, and door openings with statistical variance models (mean ± 2σ thresholds)
  • Implement adaptive threshold adjustment where detection sensitivity automatically tightens during stable operation periods (nighttime, low-activity hours) to 5g/year equivalent and relaxes to 15g/year during high-noise windows (defrost, maintenance), with real-time pattern matching against the learned signature library to suppress known benign transients
  • Integrate operator feedback loop where technicians classify each alarm as true/false via mobile interface, triggering automatic recalibration of detection parameters within 24 hours—false alarm adjusts threshold +10% for that event type, confirmed leak tightens sensitivity -5% system-wide, converging to 95%+ accuracy within 30 days of deployment
Expected Effect : False alarm rate <5%, leak detection 5-10g/year, 95%+ accuracy in 30 days, no hardware addition
Risk Control :
  • initial training period requires leak-free operation assumption
  • algorithm requires minimum 500 data points for statistical validity
  • operator feedback compliance below 60% degrades learning effectiveness

Problem Direction 3 :

ImproveDetection response time
VS
ConstraintEquipment system complexity

Inspiration 1 : Cross-domain reference

Application Principle: #19 Periodic action
Cross-domain applicability Assess applicability
Sand separator interface detection
Innovative Solution Refine solution

Pulsed refrigerant pressure interrogation for rapid leak detection

Apply controlled pressure pulses to detect leaks without continuous monitoring
How to solve :
  • Inject brief pressure pulses (±0.2 bar, 10-second duration) into refrigerant circuit every 15 minutes using existing compressor modulation
  • measure pressure decay rate over subsequent 60-second window using standard pressure transducer
  • leak rates ≥5g/year produce decay slopes >0.015 bar/min versus <0.003 bar/min for intact systems
  • Implement single-chip microcontroller (e.g., STM32 series) performing local slope calculation and threshold comparison
  • transmit only binary leak/no-leak status via low-power wireless (LoRa, 2.4 kbps), eliminating continuous data streaming and central processing infrastructure
  • Establish baseline decay profile during 48-hour commissioning period across three temperature zones (ambient, operating, defrost)
  • system auto-adjusts thresholds ±20% per zone, filtering normal fluctuations while maintaining 5-10g/year sensitivity with <5% false alarm rate
Expected Effect : Response time 15 min, infrastructure complexity -70%, detection sensitivity 5g/year, false alarm <5%
Risk Control :
  • pressure pulse interference with cooling cycle
  • transducer drift affecting decay measurement
  • baseline variation across seasonal temperature changes

Problem Direction 4 :

ImproveDetection response time
VS
ConstraintOperational false alarm rate

Inspiration 1 : Cross-domain reference

Application Principle: #10 Preliminary action
Cross-domain applicability Assess applicability
Medical procedure monitoring system
Innovative Solution Refine solution

Adaptive baseline learning system for refrigerant leak detection with transient event pre-filtering

System learns normal operational patterns before activating alarms
How to solve :
  • Deploy 2-4 week commissioning phase where sensors record all pressure, temperature, and flow variations during defrost cycles, door openings, compressor starts, and ambient changes — building a transient event signature library with 200+ cataloged patterns per unit
  • Implement dual-stage verification algorithm requiring anomalies to persist for 5-10 consecutive minutes and match neither the learned transient library nor current operational state (defrost active, door open) before triggering alarms — transient events resolve within 2-3 minutes while 5-10g/year leaks persist continuously
  • Integrate operator feedback loop where marked false alarms automatically adjust unit-specific detection thresholds within ±15% and add new transient patterns to the library, converging to optimal sensitivity within 7-14 days of deployment
Expected Effect : False alarm rate reduced 90%, 5g/year leak detection within 8-12 minutes, 95%+ accuracy after commissioning
Risk Control :
  • insufficient commissioning data diversity
  • operator feedback inconsistency
  • baseline drift from seasonal changes

Problem Direction 5 :

ImproveLeak identification accuracy
VS
ConstraintEquipment system complexity

Inspiration 1 : Cross-domain reference

Application Principle: #28 Mechanics substitution
Cross-domain applicability Assess applicability
Manipulator system
Innovative Solution Refine solution

Refrigerant-selective infrared absorption spectroscopy for leak detection

Replace multi-sensor arrays with single optical detection system
How to solve :
  • Deploy tuned infrared spectroscopy at refrigerant-specific absorption wavelength (e.g., ammonia at 10.3–10.8 μm, R134a at 8.5–9.0 μm) to achieve molecular-level selectivity
  • Install single-beam IR sensor with bandpass optical filter (±0.2 μm tolerance) and thermopile detector (sensitivity ≥200 V/W) scanning system perimeter at 1-minute intervals
  • Implement baseline-corrected absorption algorithm calculating ΔAbsorbance = (I₀ - I)/I₀, where leak detection threshold set at ΔA ≥ 0.002 corresponding to 5 g/year leak rate, with 3-consecutive-reading confirmation (3-minute persistence filter) to eliminate transient false positives
Expected Effect : Accuracy 98%, complexity reduction 80%, 5g/year sensitivity
Risk Control :
  • optical window contamination affecting transmittance
  • ambient temperature drift impacting detector baseline
  • refrigerant type mismatch with filter wavelength
Patsnap Eureka Solution