Machine Learning Real-time System Monitoring Software
Sensoleak’s real-time monitoring system is based on proprietary machine learning algorithms that analyze streaming data provided by the operators’ systems.
Artificial Intelligence Data Mining
Our data mining system analyzes and learns thousands of time- stamped data strings. The system analyzes this large quantity of data by using a proprietary BI engine, which also allows for further dissections and for the generation of reports.
Data Mining Data Analysis
There are several types of data related to pipeline network leakages, and to leakage event identification, such as sudden leakages, small and growing leakages, and leakages resulting from maintenance works.
SensoLeak has developed a real-time monitoring system for predicting and identifying evolving failures in various pipelines and mechanical parts (sensor agnostic). The customized diagnostic system runs parallel to the SCADA and is based on a statistical algorithm that processes data obtained from sensors that are installed on the diagnosed pipeline/mechanical part. Whenever the machine-learning algorithm receives SCADA data, it calculates the “system health grade” (HG), makes a decision concerning system normality and outputs alerts as necessary when it identifies deviations from the norm (anomalies). These alerts are very accurate and delivered to the control room operators in any form specified or required by the client. At a Glance.
Industries and Applications
Oil, Gas, and Water Pipelines – leak and spill prevention, theft, predictive maintenance
Turbines – inefficiency of power generation
Liquid pumps – cavitations and vibrations
Motors – shaft misalignment
Pharmaceutical – chemical reactions
Petrochemical – physical and chemical influences
Heavy Machinery – pre-detection of shaft malfunctions
Any industry with stationary processes (including those with low levels of stationarity)
The system reports evolving failures long before the emergence of real ones. This is done by monitoring both the diagnosed part and the set of variables that affect or are involved in its activity.
The system’s self-learning algorithm adapts to changes in the environment and does not require any human intervention.
A low rate of false alarms made possible through the use of explanatory variables.
The ability to acquire and process data from several types of sensors, such as acoustic, seismic, electromagnetic, mechanical, chemical, thermal, etc.
Scalable software – no need to install additional sensors or instruments.
Continuous operations with no need for calibration even after unreported changes in the system’s operating conditions
Less than 5% False Alarms
92% Accuracy Rate
The system reports on evolving failures long before the formation of a real failure
The system is self-adaptive to changes in the environment and doesn’t require any human intervention.
A low rate of false alarms is attained through the use of explanatory variables that describe changes in the explained variable.