Continuous acoustic partial discharge detection, insulation fault arcing, and intrusion monitoring across distribution substations using a low-BOM quad-MEMS microphone array and TinyML DSP—with zero optical cameras and 100% GDPR compliance.
Test how the quad-MEMS spatial array and TinyML convolutional network differentiate normal transformer hum from high-voltage partial discharge arcing and physical vandalism in real-time.
Source: Quad-MEMS Array @ 192 kHz Microsecond Sampling
{
"node_id": "ASE-SUB-DE-EON-892",
"timestamp": "2026-08-13T12:00:00Z",
"event": "NOMINAL_MONITORING",
"acoustic_pd_detected": false,
"doa_azimuth_deg": 142.5,
"confidence": 0.984,
"telemetry_size_bytes": 142,
"gdpr_compliance_status": "PASSED_HARDWARE_DSP_FILTER"
}
How constraint-driven engineering at IIT Madras eliminated 90%+ of traditional utility thermal surveillance hardware costs while adding spatial acoustic intelligence.
Arranged in a 50mm cross geometry to capture acoustic phase differences across microsecond delays. Capable of directional spatial filtering up to 80 kHz ultrasonic frequencies.
Rather than streaming costly raw audio or HD optical video over cellular networks, 100% of spatial delay-and-sum beamforming and TinyML feature extraction runs in 20ms cycles on the microcontroller. Only lightweight SCADA events are transmitted.
The quad-microphone array measures inter-aural time differences (ITD) to pinpoint the precise angle of an arcing spark or mechanical impact within ±5° accuracy, identifying exact transformer bushings or switchgear bays.
High-coercivity N52 neodymium magnets clamp directly onto transformer casing walls in under 3 minutes. No outages, zero wiring, and zero electrical safety shutdowns needed for setup.
Inspect distributed EchoGrid nodes across an active regional distribution network. Select nodes to review localized acoustic health, discharge telemetry, and predictive maintenance alerts.
Adjust deployment parameters for your regional distribution network to simulate upfront CAPEX, recurring OPEX, and net financial payback.
Ensuring zero regulatory friction across European Distribution System Operators (DSOs).
Privacy-by-design. Hardware DSP filters strip human speech harmonics prior to feature quantization. Zero audio recordings or identifiable sound are ever stored or transmitted.
Hardens physical perimeter security for critical energy infrastructure. Hardware Secure Element (ATECC608) with encrypted TLS 1.3 backhaul protection.
Fully aligned with international partial discharge measurement protocols. Acoustic telemetry events map directly into utility SCADA GOOSE messages.
Fully deterministic edge inference models with transparent risk score boundaries. Zero black-box generative AI deployed on safety-critical power grid assets.
A structured co-innovation strategy spanning IIT Madras prototyping to European grid scale-up.
Hardware optimization and acoustic partial discharge dataset creation at IIT Madras high-voltage testing facility.
Deploy 200 pilot nodes on distribution transformers in extreme temperature and high humidity tropical grid zones.
Substation deployment across German and Swedish E.ON distribution networks for European winter testing.
CE / IEC certification completion and automated manufacturing rollout supported by Infosys system integration.