Proof of Concept
We make BIG claims; they are almost too hard to believe. Let’s put your skepticism on the line and you be the judge, not us.
We will do a POC with your data using AI-MSET with 100,000+ signals, high sampling rates (even sub-second or kHz-MHz rates)(yes, vibration and acoustic transducers too), spanning several years or more using your actual data. With the results, you compare our output with any tool available in the marketplace today or internally developed algorithm(s) you are using today.
We will demonstrate, quantitatively:
- Earliest time to detect an anomaly (weeks to months vs. minutes to hours)
- Extremely low False Alarms (non-threshold based, probability 1:10^5 over 10,000 hours). Probability tolerance set in advance of training and independent to the Missed Alarm probability.
- Extremely low Missed Alarms (non-threshold based, probability 1:10^5 over 10,000 hours). Probability tolerance set in advance of training and independent to the False Alarm probability.
- Resources required (both compute and human) will be 2-3 orders of magnitude less than conventual approaches.
- Effort to maintain the model, i.e., replace or added sensors, model expansion, etc.
Then, you be the judge.