OTOSENSE
AI SOFTWARE
Advanced Predictive Maintenance Software
OtoSense AI software provides more than just an algorithm: it integrates into a platform that enables a scalable and production-ready architecture. Designed to be an ally of your technicians and operators, it helps them direct their attention and effort to the machines that need it most.
It provides great value from day 1 by automatically creating an outlier model without any human help. Just a few hours after data collection begins, you can be warned of anomalies coming from your machine.
The full power of the OtoSense AI software is realized by exploring data and tagging it (labeling it). In this way, your in-house machine experts transfer their in-depth knowledge, gained over time, to the OtoSense AI.
|
|
SENSE
• Sense
• Digitize
• Featurize
|
|
BUILD
• Explore data
• Create AI models
• Accumulate knowledge
|
|
INTERPRET
• Detect anomalies
• Add domain expertise
• Fuse sensors
|
|
EXTRACT VALUE
• Store
• Aggregate
• Notify
• Build services
The OtoSense AI Software is Modular and Flexible
The predictive maintenance software provides a full range of turnkey solutions. Each component, from sensing to data interpretation, can be used independently, including the data acquisition module, featurization, and visualization dashboards. Use the entire package or design a tailor-made solution–it’s up to you. OtoSense modules easily integrate with your existing solution.
The OtoSense AI Software is Easy to Deploy
Developed as software containers—OtoSense is adaptable to any implementation, enabling the software to run anywhere. By default, the predictive maintenance software is hosted on AWS. However, the turnkey solution can run on almost any enterprise cloud or on-premise in any environment. The OtoSense AI software can even run on a physical server, hosted on a computer, or your premises—creating a fully self-contained system. Whatever your company’s data storage and security guidelines, there’s an easy way to install OtoSense and take advantage of the AI software’s many benefits.
Explore Your Data
Unsupervised Mapping
This “out of the box,” machine learning tool automatically maps a selection of sounds or vibrations by similarity. Selections can be “all the data collected so far” on an asset or by specific time-range. Once the data is collected, OtoSense selects a few hundred data points that best represent the full variety of sounds or vibrations and then maps them on screen. A technician or engineer, using the exploration interface, can name those points and later search by tag or similarity.
Listen & Tag
Requires human intervention to fine-tune and add value by making selections, listening to data, and tagging. This feature provides for easy interaction between OtoSense’s machine learning capabilities and your domain specialist. Listen & Tag enables an exploration of the data collected by time, using the outlier score as an entry point. The tool also offers a visualization of the waveform and the spectrogram of the data while listening. Your engineer or technician can make a precise selection on the timeline and tag (name) the raw data. Each tag constitutes a new event associated with the asset. Any new behavior is tagged in this way (for example, “cavitation,” “leaky_piston” or “worn_seal”). The more events listed, the more precise the diagnostic capabilities of the tool, and the more valuable in detecting or predicting incidents.
Build and Test Your Models
Anomaly (Outlier) Detection Model
By default, OtoSense machine learning automatically generates and applies a first outlier detection model, one-hour after data collection begins, then fine-tunes it for a week. Once applied, the model continuously calculates, in real-time, the outlier score of the sounds or vibrations collected. When outlier scores reach a defined threshold, notifications or alerts indicating an anomaly are sent.
An anomaly detection model can also be built manually, with the aid of your technician or engineer, by selecting data that best references the normal signal of the asset (for example, the constant hum of the motor at different engine speeds).
Event Recognition Models
This tool allows OtoSense to interpret collected signals in real-time, giving the raw data meaning. Events, from minor hiccups to the early indications of an impending breakdown, are first identified and defined by a human using the exploration tool. Then, OtoSense’s intuitive interface is employed to build models and teach the system to recognize both the subtle and ominous signs of issues, elevating OtoSense’s predictive maintenance software to a valuable diagnostic tool.
Try Our No-Cost OtoSense Trial
Quickly evaluate OtoSense using sound or vibration samples from your assets. Collect a few minutes of data on your asset, upload it, and simply build a model. Then, check how OtoSense interprets your data across the user interface.
