Meyd675 -
| Element | Description | |---------|-------------| | Health Bar Grid | 4‑column card per asset. Green = Healthy, Yellow = Degrading, Red = Critical. Hover shows current KPI values (OEE, Energy %, RUL). | | Global KPI Strip | Real‑time OEE, Energy Consumption, #Active Alerts. | | Quick Filters | By line, location, alarm severity. | | “Drill‑Down” button | Opens the Asset Detail page (single click). |
The MEYD‑675 is a high‑performance, low‑power System‑on‑Chip (SoC) designed specifically for edge‑AI workloads in industrial, automotive, and consumer‑grade devices. By combining a heterogeneous compute fabric with an on‑die AI‑optimized memory subsystem, the MEYD‑675 delivers up to 2 TOPS/W (tera‑operations per second per watt) while maintaining a compact 12 mm × 12 mm footprint in a 7 nm FinFET process.
Key selling points:
| Feature | Benefit | |---------|----------| | Hybrid Compute Engine – 4× ARM Cortex‑A78AE + 8× custom AI‑matrix cores | Seamless handling of control‑plane code and massive data‑parallel inference | | Unified 8 GB LPDDR5X on‑die with 2 TB/s bandwidth | Eliminates off‑chip memory bottlenecks, reduces latency | | Integrated Secure Enclave (TEE) | Hardware‑rooted attestation, secure model deployment | | Dynamic Voltage & Frequency Scaling (DVFS) + power islands | Fine‑grained power management for battery‑operated devices | | Standardized I/O – PCIe 4.0 x4, USB 3.2, MIPI‑CSI/DSI, Ethernet 1 GbE | Easy integration into existing hardware ecosystems | | Software Stack – Open‑source SDK, ONNX runtime, TensorFlow‑Lite micro | Fast time‑to‑market for developers |
| NFR‑ID | Description | Target | |--------|-------------|--------| | NFR‑001 | Latency – End‑to‑end detection (sensor → alert) ≤ 250 ms for high‑frequency streams. | 250 ms | | NFR‑002 | Resource Footprint – ≤ 300 MB RAM, ≤ 1 W CPU on MEYD‑675 ARM‑Cortex‑A53. | 300 MB / 1 W | | NFR‑003 | Scalability – One hub can manage up to 200 sensors; horizontally scale to thousands of hubs via Kubernetes at the cloud tier. | 200 sensors/hub | | NFR‑004 | Reliability – 99.9 % uptime for the edge runtime; automated watchdog restart. | 99.9 % | | NFR‑005 | Data Retention – Raw sensor data kept locally for 48 h; aggregated metrics persisted 90 days in cloud. | 48 h / 90 days | | NFR‑006 | Usability – Dashboard onboarding < 15 min; “Explain‑Why” drill‑down ≤ 2 clicks. | 15 min / 2 clicks | | NFR‑007 | Compliance – GDPR‑compatible data handling, optional anonymisation of device IDs. | GDPR‑ready | | NFR‑008 | Maintainability – All edge components containerised; CI/CD pipeline with automated regression testing (≥ 90 % code coverage). | CI/CD ready | meyd675
| Block | Description | |-------|-------------| | ARM Cortex‑A78AE (4‑core) | General‑purpose cores with advanced reliability extensions (A‑R‑E) for real‑time control, OS, and pre‑/post‑processing. | | AI Matrix Cores (8×) | 64‑bit fixed‑point MAC arrays, each with 256 KB local SRAM, supporting INT8/INT16 and mixed‑precision FP16/FP32. | | DSP Subsystem | 2× 32‑bit VLIW DSPs for audio/vision signal processing, complementing matrix cores for non‑tensor workloads. | | RISC‑V Security Coprocessor | Handles cryptographic primitives, secure boot, and key management. |
The matrix cores are tightly coupled to the on‑die memory via a crossbar that guarantees ≤ 15 ns latency for data fetches, enabling model‑in‑memory execution without external DRAM stalls. | Element | Description | |---------|-------------| | Health
| # | As a … | I want … | So that … | |---|--------|----------|-----------| | 1 | Operator | to see a single, colour‑coded health bar for each critical asset on my HMI | I can instantly spot which machine needs attention without digging through logs | | 2 | Maintenance Engineer | an auto‑generated RCA notebook when an alarm fires (including sensor traces, correlation graphs, and probable cause) | I spend minutes, not hours, fixing the issue | | 3 | Production Planner | a predictive output forecast for the next 24 h based on current equipment health and process set‑points | I can adjust shift plans and inventory proactively | | 4 | Business Analyst | a monthly “Insight Dashboard” that aggregates OEE, energy usage, and anomaly trends across all MEYD‑675 hubs | I can report ROI and justify further automation investments | | 5 | IT/DevOps | a plug‑and‑play container that can be deployed on the MEYD‑675 edge runtime (Docker‑Slim) | I avoid complex installs and can roll out updates centrally |
| Aspect | Description |
|--------|-------------|
| Name | Adaptive Insight Engine (AIE) – “MEYD‑675 Insight Layer” |
| Goal | Transform high‑frequency sensor data from MEYD‑675 into real‑time, context‑aware recommendations, anomaly‑driven alerts, and predictive maintenance schedules without requiring a data‑science expert on‑site. |
| Primary Users | • Plant floor operators
• Maintenance engineers
• Production planners
• Business analysts / executives |
| Business Value | • 10‑20 % reduction in unplanned downtime
• 5‑8 % increase in overall equipment effectiveness (OEE)
• Faster root‑cause analysis (RCA) → lower labor cost
• Ability to monetize data (trend reports, compliance dashboards) |
| Key Differentiators | 1️⃣ Edge‑first analytics (no need for constant cloud round‑trip)
2️⃣ Self‑learning models that auto‑tune to each plant’s unique operating envelope
3️⃣ “Explain‑Why” UI that surfaces sensor‑level evidence for every recommendation | | Block | Description | |-------|-------------| | ARM
