Industrial telemetry & predictive maintenance
Plan maintenance around the machine, not the calendar.
Parallax collects telemetry from PLCs, SCADA systems and sensors through an edge gateway that keeps buffering when the network drops. It fuses those signals into a live model of each asset, detects multivariate anomalies, estimates remaining useful life and measures OEE line by line. Maintenance and production work from the same evidence.
VARDA ParallaxRuns on Orbit
The problem
The signals exist. They are just in too many places.
A typical plant runs controllers from several vendors, a historian or two, spreadsheets for OEE and a maintenance system that learns about a failure after it has happened. Each holds part of the story, and none connects a rising vibration trend to the order that will ship late because of it.
- 01
Controllers and protocols from several vendors on the same line
- 02
Unplanned stops noticed by operators before the data shows them
- 03
Calendar-based maintenance that replaces healthy parts and misses failing ones
- 04
OEE calculated by hand, days late and differently on every line
- Line 2
- Press P-204
- Main motor M1
- Bearing B2
- Hydraulic unit H1
- Robot cell R3
- Conveyor C7
Suggested: replace bearing B2 · planned stop window Tue 06:00
- Availability
- 91,4%
- Performance
- 86,9%
- Quality
- 99,2%
- OEE
- 78,8%
Capabilities
See the machine from every angle.
Industrial connectivity
Native OPC UA, MQTT with Sparkplug B, Modbus TCP/RTU and S7 connectors read from PLCs and SCADA without touching control logic. Collection is read-only by default.
Edge gateway with store-and-forward
A hardened gateway on the plant network buffers data locally when the uplink fails and forwards it in order once the link returns. An outage leaves a delay, not a gap.
Time-series store and asset model
High-resolution telemetry is kept alongside downsampled rollups and organised in an ISA-95-style hierarchy of site, area, line, cell and asset. Live dashboards read from the same store.
Multivariate anomaly detection
Models learn how vibration, temperature, current and pressure move together on each asset, and flag departures that no single threshold would catch. Every alert shows the signals that drove it.
Remaining-useful-life estimates
Degradation models estimate how long a component can run before it needs intervention, as a range with a stated confidence rather than a single date, so work can be scheduled into planned stops.
OEE and alarm management
Availability, performance and quality are calculated from machine states and counts, with stop reasons captured at the line. Alarms are deduplicated, prioritised and routed, so the one that matters is not lost among hundreds.
How it works
From the PLC to the work order
Connect at the edge
The gateway is installed on the OT network and reads tags over OPC UA, MQTT, Modbus or S7. Control logic is untouched, and data crosses from OT to IT through a single, defined conduit.
Model the plant
Tags are mapped to the asset hierarchy with units, ranges and context such as product, shift and recipe, so that a reading means the same thing everywhere.
Learn normal behaviour
For each asset class, models learn how signals behave across operating modes, using historian data where it exists.
Detect and estimate
Live data is scored continuously. Anomalies are raised with the signals behind them, and remaining-life estimates are updated as wear progresses.
Act and feed back
Confirmed alerts become work orders in your maintenance system. Technicians’ findings are recorded against the alert and used to refine the models.
Architecture
VARDA Parallax · Architecture
01 Sources
- PLCs & controllers
- SCADA & historians
- IIoT sensors
- MES & ERP
02 Processing
- Edge gateway
- Enterprise connectors
03 VARDA Parallax
- Asset model & time series
- Anomaly, RUL & OEE analytics
04 Outputs
- Dashboards & OEE reports
- Alerts & alarm routing
- Work orders (CMMS/EAM)
Use cases
Use cases
Motors, pumps, fans and gearboxes are monitored through vibration, temperature and current. Parallax flags bearing wear, misalignment and imbalance early enough to plan the repair into a scheduled stop.
- Vibration spectra and envelope analysis computed at the edge
- Remaining-life ranges for bearings and gearboxes
- Work orders raised with the evidence attached
Machine states, counts and reject signals become availability, performance and quality per line and shift. Stops are classified at the source, so the loss tree reflects what actually happened on the floor.
- OEE derived from PLC states, not manual logs
- Short stops and speed losses measured, not estimated
- Shift reports comparable across lines and sites
Compressors, chillers, boilers and main meters are tracked against production output. Parallax identifies idle consumption and compressed-air leaks, and attributes energy to each unit produced.
- Energy per unit by line, product and shift
- Base-load alerts outside production hours
- Compressor and chiller anomalies caught before they become breakdowns
Deployment options
Deployment options
On-premises
Everything runs inside the plant: gateways on the OT network, the time-series store and analytics on servers in your own server room. Suited to sites with strict OT segmentation or limited connectivity.
Private cloud
Gateways stay at each site and forward data over an encrypted, outbound-only connection to Parallax in your private cloud. Suited to groups with several plants that want one fleet view and shared models.
Hybrid
Anomaly detection and alarms run at the edge, so monitoring continues when the uplink is down; history, fleet analytics and model training run centrally.
Technical specification
VARDA Parallax
- Protocols
- OPC UA, MQTT / Sparkplug B, Modbus TCP/RTU, S7, EtherNet/IP
- Edge buffering
- Local store-and-forward, sized for outages of several days (design target)
- Sampling
- Sub-second to minutes per tag; deadband and report-by-exception
- Time-series storage
- Raw data plus rollups; retention and downsampling per tier
- Analytics
- Multivariate anomaly detection, RUL models, spectral features
- Asset model
- ISA-95-style hierarchy; tag templates per asset class
- OEE model
- Availability × performance × quality; configurable loss tree
- Security
- Outbound-only connections, TLS 1.3, certificate-based device identity
Integrations
- OPC UA
- MQTT
- Sparkplug B
- Modbus TCP/RTU
- Siemens S7
- EtherNet/IP
- BACnet
- AVEVA PI System
- Apache Kafka
- SAP PM
- IBM Maximo
- Grafana
- REST API
Compliance & governance
- Designed to support IEC 62443 practice: zones and conduits, least-privilege access and outbound-only connections from the OT network.
- Designed to support ISO/IEC 27001-style controls for access management, encryption in transit and at rest, and audit logging.
- Designed to support KVKK and GDPR where telemetry is linked to people, such as operator actions and shift records, through role-based access and configurable retention.
- Designed to support ISO 50001 energy reviews with metered consumption per line, product and shift.
Frequently asked questions
Frequently asked questions
01Will connecting Parallax affect our PLCs or the control network?
Collection is read-only by default and uses interfaces your controllers already expose, such as OPC UA servers or Modbus registers. Polling rates are agreed with your automation team and tested before go-live. Parallax never writes to a controller unless you explicitly enable and approve a specific write path.
02Our machines are old and have few sensors. Can Parallax still help?
Often, yes. Many older machines already expose useful signals through their PLC, such as states, counts and motor current, which is enough for OEE and basic health monitoring. Where condition data is missing, vibration and temperature sensors can be retrofitted without touching the control system, and we help decide which assets justify it.
03How much history do the models need before they are useful?
Anomaly detection can start from a few weeks of normal operation, provided that period covers the main operating modes. Remaining-useful-life estimates need examples of degradation or failure, so they improve as maintenance outcomes are recorded. Until then, we show confidence ranges rather than a precise date.
04What happens when the link between the plant and the data centre goes down?
The edge gateway keeps collecting, stores data locally and forwards it in order, with original timestamps, when the link returns. In hybrid deployments, anomaly detection and alarms continue at the edge throughout the outage. Buffer duration depends on disk size and data rate, and is sized during design.
05Does Parallax replace our SCADA, historian or CMMS?
No. SCADA keeps running the process and your CMMS keeps managing maintenance work; Parallax reads from the first, writes work orders into the second and adds the analytics layer between them. An existing historian can remain the system of record, with Parallax reading from it.
See VARDA Parallax with your own data.
Connects PLCs, SCADA and sensors over OPC UA, MQTT and Modbus, and turns telemetry into anomaly alerts, remaining-useful-life estimates and OEE.