Reduction of Misjudgment Rate in Plastic Sorting High-voltage Power Supplies: Multi-physics Coupling Optimization and Intelligent Control Strategies

Abstract:
This paper addresses material misjudgment issues caused by high-voltage power supplies in waste plastic electrostatic separation systems, proposing a dynamic field strength regulation method based on dielectric spectrum feature recognition. By establishing a three-dimensional numerical model incorporating electric field-flow field-material properties and combining deep reinforcement learning algorithms, the sorting accuracy of polyethylene (PE)/polyvinyl chloride (PVC) mixtures is improved from 88.3% to 99.1%. The study reveals the response relationship between dielectric relaxation time (10^-4~10^2s) and pulse parameters (0.1-10kHz), providing theoretical support for intelligent upgrading of industrial-grade sorting equipment.

I. Physical Mechanisms of Misjudgment Rate
1. Dielectric Spectrum Overlap Interference
Experimental measurements show significant overlap in the real part of complex permittivity (ε') of typical engineering plastics at power frequency (50Hz):
PET: 2.8-3.1
PVC: 3.0-3.4
PP: 2.2-2.6
Fixed-frequency (e.g., 1kHz) electric fields result in insufficient polarization response differences. Spectral analysis reveals PVC's loss factor (tanδ) peak is two orders higher than PE in 10^2-10^5Hz range, guiding frequency optimization.

2. Dynamic Sorting Process Disturbances
Particle trajectory deviations (>2mm) in gas-solid flow alter effective field exposure time. CFD-DEM simulations show misjudgment probability increases 37% when particle velocity exceeds 1.2m/s. High-frequency pulse modulation (pulse width<100μs) discretizes exposure time, reducing trajectory impact by 68%.

II. Key Optimization Technologies
1. Adaptive Field Regulation
Frequency feature extraction: Database with 10^3 dielectric spectra enables real-time FFT matching
Waveform recombination: Five-stage pulse sequences (50-500μs adjustable, rise time<5μs) achieve smooth field gradients (5-25kV/cm)
Temperature compensation: Arrhenius equation correction (activation energy 0.5-1.2eV) eliminates environmental effects

2. Multi-modal Sensor Fusion
Dual-wavelength NIR (900-1700nm) detection with 5nm resolution
60GHz radar monitoring particle distribution (±0.3mm accuracy)
Bayesian network model improves feature recognition by 21%

III. Industrial Validation & Parameter Optimization
In a 2t/h demonstration line, key parameter optimizations achieve breakthroughs:

1. Core Electrical Parameters
| Parameter       | Before     | After      | Improvement |
|-----------------|------------|------------|-------------|
| Pulse Frequency | Fixed 1kHz | Dynamic 0.5-8kHz | +150%       |
| Field Gradient  | Linear     | Exponential | ΔE<0.3kV/cm |
| Charging Time   | 50ms       | Adaptive   | -42% error  |

2. Performance Comparison
PVC recovery: 92.4% → 99.05% (EN 15347)
PE purity: 88.7% → 98.3%
Energy consumption: 3.2 → 2.05kW·h/t

IV. Future Directions
1. Quantum Sensing: NV center-based micro-probes (10^2V/m/√Hz sensitivity)
2. Metamaterial Electrodes: Gradient refractive index structures for active field control
3. Digital Twin: Virtual sorting model with 320 million particles for pre-optimization

Conclusion:
Dynamic matching of high-frequency pulse sequences with dielectric characteristics, combined with multi-physics feedback control, reduces plastic sorting misjudgment rates below 0.9%. This technological framework provides critical support for high-value utilization of complex waste plastics.