Predictive maintenance
Models that learn the normal behavior of each asset and identify deviations that precede failures, so interventions can be planned at the right time.
Artificial intelligence applied to industry
Anticipate failures.
Decide intelligently.
Akatá develops intelligent decision-support and predictive maintenance systems that turn your equipment data into early, explainable alerts.
Portfolio
Solutions
Artificial intelligence built on real operational data, so teams can act before problems happen.
Decision support
Intelligent systems that organize signals, context and explanations so managers and operators can decide more safely and quickly.
Intelligent equipment monitoring
Our main product: Akatá Marandu monitors sensors in real time and issues early anomaly alerts, each one with its explanation.
See the productProduct
Akatá Marandu
An equipment monitoring platform that combines sensor data and machine learning models to detect anomalies and warn operations in advance.
In Old Tupi, Marandu means news, notice: information that arrives before the event.
Equipment A
Warning
Top contributions to the alert
- Lube oil temperature
- Winding temperature
- Jacket temperature
-
Continuous monitoring
Tracks dozens of sensors per piece of equipment, such as temperatures, pressures, electrical quantities and vibration, even when they arrive at different sampling rates.
-
One model per asset
Each piece of equipment has its own data processing, feature engineering and a model trained on its real operating history.
-
Reliable alerts
Score smoothing, a dynamic threshold, a persistence requirement and a re-arm period prevent alarms triggered by isolated spikes.
-
Explainable alerts
Every alert shows the variables that contributed most to it, with the observed value and the weight of each one, to guide inspection.
-
Operations dashboard
Equipment status, sensor history, anomalies by period, event reports, notifications and history export.
-
Extensible, secure architecture
New equipment is added as independent modules. Models are delivered compiled, and access is protected by signed-token authentication.
How it works
-
Signal integration
Sensors with different sampling rates are unified on a common time base, preserving the information from each source.
-
Processing and features
Operating-regime filtering, data cleaning and consistency checks, and statistical features over rolling windows.
-
Modeling and validation
Training on the equipment's history and rigorous time-based validation that never mixes past and future.
-
Operational decision
An alert is issued only when the risk pattern persists, and it reaches the team together with its explanation.
About
The fire of the mind
Akatá is a name built from Brazilian Indigenous roots of the Tupi-Guarani family.
- Akã
- mind, head, intelligence
- Tatá
- fire, flame, energy
Together, they evoke a brilliant mind: an active intelligence, driven by the warmth of creativity and a passion for knowledge. That is what we seek in every project, combining ancestral wisdom, innovation and the energy to get things done. Hence our motto: energized intelligence.
Origin
Born from applied research
Akatá is a spin-off of a partnership between science and technology institutions and CENPES, Petrobras' research center. The research that gave rise to the company was developed and validated in real industrial operations.
Science and technology institutions
-
Federal University of Alagoas (UFAL)
-
Federal University of Paraíba (UFPB)
Research partnership
-
CENPES — Petrobras Research Center
Project partners
-
Petrobras
-
Transpetro
Partners
Contact
Shall we talk about your assets?
Tell us which equipment you want to monitor and which decisions you need to make with more confidence.
- Email contato@akata.tec.br