Asset Accelerator · Industry

From sensors
to two weeks
of failure headroom.

A pre‑built asset health platform — sensor adapters, edge runtime, spectral feature extractors, anomaly models, alert routing, and the operator dashboard your plant manager will actually use. Production-grade in ten weeks, not ten quarters.

Book a pilot scope
Latest: v1.8.2, Jan 2026 License: Per plant

What you actually
get on day one.

9 modules · plant-deployable
M.01

→ Sensor adapter library

42 pre-built adapters — IFM, SKF, Schneider, Siemens, generic 4-20mA. OPC-UA, Modbus, MQTT pluggable.

OPC-UA · Modbus · MQTT
M.02

→ Edge runtime

Containerised runtime for NVIDIA Jetson / industrial PCs. OTA updates with rollback. Field-recovery playbook.

Jetson · Yocto · OTA
M.03

→ Telemetry broker

MQTT Sparkplug B broker + time-series DB. Local buffer when the WAN drops. Reconcile-on-reconnect built in.

EMQX · InfluxDB
M.04

→ Spectral feature extractors

FFT, envelope, crest factor, kurtosis. Per-asset baselines learned, drift tracked. The features your model needs to be honest.

librosa · NumPy
M.05

→ Anomaly + failure models

Three baseline models per asset class — unsupervised anomaly, supervised failure classifier, remaining-useful-life regressor.

PyTorch · XGBoost · ONNX
M.06

→ Asset registry

Every asset modelled with hierarchy (plant → line → unit → component). Per-asset model assignment, calibration, history.

Postgres · ISA-95
M.07

→ Alert routing

Priority queues. Work order auto-creation in your CMMS. Integration with SAP PM, Maximo, Fiix, UpKeep.

SAP PM · Maximo · Fiix
M.08

→ Operator dashboard

Designed for plant managers, not data scientists. Six-second read. One thumb. Works in PPE gloves. Tested with actual operators.

React · Tablet-first
M.09

→ Runbook + SLA

19 named runbooks. False-positive review workflow. Plant-team training materials in English + 4 languages.

Runbook v1.8 · 5 languages

Edge to dashboard,
flattened out.

5 stages · offline-first

Five stages, two stores, offline-resilient.

asset-accelerator v1.8.2
SENSOR VIB / TEMP / CUR STAGE 1 Adapt + Buffer edge · 4ms STAGE 2 Extract Features edge · 18ms STAGE 3 Score Anomaly edge · 32ms STAGE 4 Decide Alert cloud · 14ms STAGE 5 Route to CMMS work order ↳ runs on edge (offline-capable) telemetry log · all features stored
Stage 1
Adapt + Buffer
Sensor adapter normalises any of 42 sensor types. Local ring buffer when WAN drops.
OPC-UA · Modbus
Stage 2
Extract Features
Per-asset spectral features. FFT, envelope, crest factor, kurtosis — at line rate.
librosa · NumPy
Stage 3
Score Anomaly
ONNX-compiled model on Jetson. Per-asset model assignment, calibration tracked.
ONNX · TensorRT
Stage 4
Decide Alert
Cloud-side decisioning: threshold + recent history + work-order context. Suppress duplicates.
Postgres · Rules
Stage 5
Route to CMMS
Work order auto-created in your CMMS with severity, asset, and recommended action.
SAP PM · Maximo

The ten‑week plan.
Pinned, not promised.

10 weeks · 4 engineers + 1 on-site
Week ↓Ship
Week 1
Plant walk + asset registry.
Two engineers on the plant floor with your controls and maintenance leads. Asset hierarchy, sensor inventory, CMMS audit. Critical assets prioritised.
Asset registry v1
Week 2-3
Edge install + telemetry.
Edge appliances installed. Sensor adapters wired. Telemetry flowing into the broker. Operator dashboard live in read-only mode.
Sensors live
Week 4-5
Baseline + feature extraction.
14-day baseline per asset. Spectral features extracted and stored. Unsupervised anomaly model warming up. Initial false-positive review with your maintenance team.
Baseline learned
Week 6-7
Failure-mode tuning.
Historical failure data ingested. Supervised failure classifier trained per asset class. Alert thresholds tuned with maintenance lead sign-off.
Tuned thresholds
Week 8-9
CMMS integration + shadow.
Work-order creation wired into your CMMS. Two-week shadow mode — alerts visible to maintenance but not auto-ticketing.
Shadow live
Week 10
Production cutover.
Cutover to live alerting + auto-WO. Maintenance team trained. Runbooks signed. 30-day safety-net begins.
In production

What our last
three customers said.

Verified · 2025—2026
"
Two previous IoT vendors gave us dashboards no one used. Emorphis spent week 1 on the plant floor watching our maintenance round, then built an operator UI that survives PPE gloves and a six-second glance. Adoption was immediate.
Head of Maintenance, Crompton
4 plants · 1,840 motors · live since Q4 2025
Downtime cut
38%
Lead time on faults
14days median
False positives
<5%

A plant that's
instrumented but blind?

↳ Direct line · Asset Accelerator

Tell us your sensor inventory, your asset classes, and your CMMS. We'll send a pilot scope, candidate architecture, and fixed-fee quote within three business days.

Get in Touch

Questions we
get every week.

Click to expand ↓
Q.01
Do we need to replace our sensors?+
Almost never. The 42 pre-built adapters cover most of what's already on the plant floor — IFM, SKF, Schneider, Siemens, generic 4-20mA. We'll inventory yours in week 1 and tell you if anything needs adding.
Q.02
What if our plant network is brittle?+
We assume it is. The edge runtime buffers locally and reconciles on reconnect. Six of our six production deployments have had multi-day WAN outages with zero data loss.
Q.03
How do you handle false positives?+
Tightly. Every alert your maintenance team marks "false positive" feeds a retraining queue. Median false-positive rate at week 12 across our deployments is under 5%. We monitor it weekly and tune.
Q.04
Will our PLC engineers need to do anything?+
Minimal — usually some OPC-UA tag exposure. We meet your controls engineers in their language (ladder, ST, tag namespace) and never touch a safety loop.
Q.05
Which CMMS systems do you integrate with?+
SAP PM, IBM Maximo, Fiix, UpKeep, Maintenance Connection out of the box. Generic REST + ERP-bus adapters for the rest. Three of our six customers use custom legacy CMMS.
Q.06
Can it run fully on-prem?+
Yes. The edge runtime is required to be on-prem; the cloud-side decisioning can run in your private VPC or on a K8s cluster on-prem. Two of six customers run fully air-gapped.