Emerson and SiMa.ai bring “Physical AI” to the industrial edge

Emerson announced a collaboration with SiMa.ai to embed SiMa’s MLSoC edge AI accelerator into next-generation rugged industrial PCs, enabling multi-modal AI (vision, audio, sensor data) to run directly alongside PLCs and SCADA in harsh field environments. The joint platform targets real-time safety monitoring, computer vision quality inspection, anomaly detection, and predictive maintenance without relying on cloud round trips.

For factories, this is a concrete blueprint for collapsing separate GPU boxes, analytics servers, and safety logic into a unified IPC stack that can make closed-loop decisions in milliseconds—think inline defect removal, leak detection, or flare monitoring that adjusts processes automatically. In parallel, data/AI teams will need clean, well-labeled edge datasets to keep these models accurate over time, where platforms like Klyff can help manage image and sensor labeling before deployment.

Factory-floor takeaway: if you already standardize on Emerson IPCs, start mapping which inspection and asset-health workloads could move onto this stack over the next refresh cycle, and push vendors on how they’ll package model lifecycle management, monitoring, and rollback for OT teams—not just data scientists.[packagingtechtoday]

TTTech Nerve chosen as secure edge base for PROGNOST UP! predictive maintenance suite

PROGNOST Systems is running its UP! Detect and UP! Insight compressor monitoring applications as Docker containers on TTTech Industrial’s IEC 62443‑4‑2–certified Nerve edge platform, combining local condition monitoring with cloud-based fleet diagnostics. Nerve provides centralized device and software management, offline capability, and on-prem data control to meet Cyber Resilience Act and NIS2 requirements while still enabling AI/ML analytics on reciprocating compressors.

For plants, this is a strong reference that “predictive maintenance as a service” can coexist with strict OT security and data-sovereignty constraints; when you evaluate PdM vendors, ask explicitly what edge runtime and security certifications they rely on, and how they’ll containerize workloads at your gateways or IPCs.[tttech]

Edge AI Foundation and AWS partner on reference architectures for edge AI

The Edge AI Foundation announced that Amazon Web Services has joined as a Leadership Partner, with a focus on co-developing secure, low-latency cloud-to-edge reference architectures and supporting open-source edge AI initiatives across sectors including manufacturing. The partnership aims to standardize patterns for deploying and managing edge inference while keeping developer tooling and education aligned between cloud and edge environments.

For factory teams, expect more prescriptive AWS blueprints for running inspection and maintenance models on-site (e.g., on Greengrass or edge appliances) while orchestrating training and fleet updates from the cloud; this can shorten design cycles but reinforces the need for disciplined edge data pipelines, where tools like Klyff can help ensure labeled data flowing back to the cloud is consistent and usable.[edgeaifoundation]

Ricoh deploys Thread AI’s multimodal digital-twin platform for facilities

Ricoh is rolling out Thread AI’s facility management platform, combining multimodal AI (vision, sensors, text) with digital twin infrastructure to automate building operations and maintenance workflows. While the initial focus is commercial facilities, the stack—unifying live sensor data, 3D models, and AI agents to trigger work orders—maps closely to what many plants are trying to do for utilities, HVAC, and non-production assets.

For manufacturers, this is a signal that digital-twin-plus-AI platforms are maturing around facilities; when you look at plant-wide twins, probe vendors on how easily they can ingest OT data, represent production assets, and plug in custom PdM and quality models instead of just HVAC and space utilization.[iottechnews]

Efinix launches Titanium Edge FPGAs optimized for vision and sensor fusion at the edge

Efinix announced its Titanium Edge FPGA family, claiming 50% lower static power than its prior Titanium line, System-in-Package options with integrated HyperRAM and boot flash, and high-speed MIPI I/O up to 2.5 Gbps per lane for multi-camera and multi-sensor workloads. Devices span from 39k to 123k logic elements, with small 5.5 × 5.5 mm packages and planned security variants that integrate post-quantum cryptography, secure boot, and PUF-based keying for long-life industrial deployments.

On a line, these parts are a fit for always-on visual inspection or vibration-plus-vision fusion where you need deterministic latency, low power, and long-term availability—e.g., embedded into smart cameras or condition-monitoring nodes instead of full x86 IPCs. For teams standardizing on edge FPGAs, this is a good moment to revisit your build-vs-buy stance on PCIe cards or modules that already wrap this silicon.[efinixinc]

Swissbit previews industrial PCIe SSDs tuned for edge AI and automation

Swissbit’s Automate 2026 announcement previews two PCIe SSD series: the N7000 (4-channel, 80GB–4TB M.2) targeting reliable boot and light workloads, and the A2000 (8-channel PCIe Gen4/NVMe, 480GB–8TB, in M.2/E1.S/U.2) for low-latency, data-intensive edge AI and automation systems. Both emphasize industrial temperature ranges, vibration tolerance, power-loss protection, and long-term availability backed by in-house NAND packaging and controlled BOM.

For edge vision PCs and PdM gateways, storage stability is often overlooked—these drives are worth tracking if you’ve seen data corruption or premature SSD wear in existing IPC fleets running logging plus AI inference. Use events like Automate to push vendors on their storage qualification and lifecycle guarantees, especially for systems that must buffer high-frequency sensor data locally.[swissbit]

OnLogic to demonstrate practical Physical AI and full-facility edge rollout at Automate 2026

OnLogic announced its Automate 2026 lineup will center on “practical Physical AI”—live demos include an edge-only forklift near-miss detection system with viso.ai and a YOLOv10-based machine-vision cell that rejects rotating-part defects in real time using an online sorting solenoid. The company is also showcasing a full hardware stack: compact CL260 DIN-rail PCs, Helix and Karbon rugged boxes for line and mobile robots, high-performance Axial rack servers, and updated Karbon 800 units on Intel’s latest Core Series 2 processors.

For factories, this is a useful catalog of “right-sized” compute options for everything from small edge nodes in control cabinets up to plant-level AI servers; when planning deployments, align each workload (inspection cell, AGV, historian, PdM) with a matching hardware tier rather than defaulting to a single over- or under-provisioned box. Platforms like Klyff can then provide a common layer for packaging and monitoring models across this heterogeneous fleet.[i40today]

Where to Put Intelligence: Edge AI for Factory Maintenance — Explains how to split AI workloads between smart sensors, edge gateways, and cloud to balance latency, bandwidth, and fleet-wide learning for factory maintenance.[iiot-world]

Edge Computing in Manufacturing: Processing Data Closer to the Source — Outlines how moving analytics to the edge lets machines perform real-time anomaly detection, control actions, and data filtering, reducing latency and cloud dependency in production environments.[embeddedcomputing]

Edge Computing for Predictive Maintenance: Processing Data at the Machine — Deep-dive from iFactory on why cloud-only PdM architectures hit latency and bandwidth ceilings, and how <10 ms edge inference, local filtering, and hybrid edge–cloud workflows change ROI and rollout timelines.[ifactoryapp]

Digital Twins for Predictive Maintenance with Real-Time Virtual Asset — Explores building digital twins that stay in lockstep with live asset data to support condition monitoring, scenario simulation, and maintenance decision-making.[ifactoryapp]

Acoustic Emission and Ultrasonic Monitoring for Predictive Maintenance — Breaks down how acoustic and ultrasonic sensing can detect early-stage faults that vibration or temperature might miss, and how to integrate these channels into an edge-first PdM stack.[ifactoryapp]

AI Predictive Maintenance for Ammonia Refrigeration Compressors — A vertical deep-dive on using AI-based PdM to reduce unplanned downtime and safety risk in industrial refrigeration, with guidance that generalizes to other high-criticality rotating equipment.[ifactoryapp]

The Industrial Edge: Where Field Data Becomes Dependable — Discusses how robust edge architectures, conditioned data, and OT-aware design turn noisy field signals into trustworthy inputs for analytics and AI.[industrialautomationindia]

Industry at the Edge — Reflects on lessons from deploying an industrial edge AI platform in real projects, including integration with existing OT, condition monitoring use cases, and managing distributed edge nodes.[barbara]

Quality leaders

Maintenance & reliability

Data/AI / digital transformation

Send up details on what you would like us to cover in Manufacturing Intelligence

TWIMI is published weekly. The scope covers developments from the prior 7 days or earlier if that ties into the stories for this week. No vendor relationships influence coverage. Forward to a colleague in ops, quality, or IT/OT — the more disciplines reading from the same page, the faster deployments happen.

Team twimi