Unexpected breakdowns in a power plant can cause considerable disruption to the power grid. Real-time analysis, real-time response, and real-time control for predictive maintenance and prevention of power plants are real-time, high-volume data acquisition use cases.
Pumps in hydroelectric and thermal power plants can suffer huge damage and have a big impact on downtime.
To mitigate downtime, edge intelligence devices need to collect real-time sensor data. By analyzing the vibration spectrum, monitoring pressures as well as other machine conditions, power companies can prevent unexpected maintenance breaks.
power plants consist of super-large rotary shafts and electromagnets. In order to preserve these facilities, they need to monitor and discovered abnormal symptoms of pumps and motors such as foreign product detection. However, customer costs have far exceed initial expectations primarily due to increased maintenance costs and lower diagnostic accuracy.
As a solution, nondestructive testing and facility diagnostics needs to be improved. It is necessary to monitor rotating bodies and perform vibration spectrum analysis at increased data precision.
The data collection of the existing facilities are in minutes! and in the case of vibration data over 15 minute periods.
However, data must be collected in 0.1 second increments to capture fine initial defects.
Equipment vibration is measured via spectral analysis, and then pre-processing analysis with historical data verified with machine learning. Analytical data should be analyzed through artificial intelligence algorithms to provide specific implementation methods for facility calibration and maintenance.
Effect of application
By building a “Data Lake”, which stores data for deep learning, Hadoop was inadequate. A time series DBMS is better suited and results in saving in learning and maintenance costs. Machbase’s time series database was implemented in order to build an optimized defensive strategy. Machbase was used to perform data acquisition along with statistical analysis of new defect types in 0.1 second increments, and to perform association analysis comparing
each facility against each other.