Compressor Sensor Failure Estimation via Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Gas compression systems with variable speed electric motors face challenges in maintaining high availability and reliability due to compressor surge events, which can lead to production delays and economic losses, especially when sensor failures occur, causing abrupt shut-downs.
Innovation Solution
A method and system that utilize electrical parameters from the drive and motor to estimate and validate process variables, allowing for continued operation by generating control signals even in the absence or failure of sensor signals, thereby maintaining production and increasing availability and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor measurements are used to monitor and control the compressor, then measurement precision is improved, but reliability deteriorates when sensor failures occur causing abrupt shut-downs
Solution Approach 1:
The patent creates a virtual copy of the failed sensor measurement by using a neural network model that estimates the process variable (such as compressor surge margin) based on electrical parameters from the motor and drive system. This virtual measurement replaces the failed physical sensor signal, allowing the control system to continue operating without abrupt shut-downs while maintaining the functional capability that the original sensor provided.
Solution Approach 2:
The patent introduces an intermediary estimation model (neural network) that mediates between the available electrical parameters and the required process variable measurements. When a sensor fails, this intermediary model bridges the gap by providing estimated values that allow the control system to maintain operation, thus improving reliability while preserving measurement functionality.
2Reliability
If anti-surge control systems with recycle valves are implemented, then compressor protection is improved, but device complexity increases
Solution Approach 1:
The patent replaces the need for complex mechanical sensor installations and associated failure modes with an electrical-based estimation system. By using electrical parameters from the motor and drive (current, voltage, frequency) to estimate process variables through a neural network model, the system achieves compressor protection functionality without the complexity and failure susceptibility of additional physical sensors and their mounting infrastructure.
3Reliability
If electrical parameters are used to estimate process variables, then reliability is improved by avoiding sensor failures, but measurement precision may deteriorate
Solution Approach 1:
The patent implements a feedback mechanism where the neural network model continuously refines its estimates of process variables based on the relationship between electrical parameters and actual process conditions. The model is trained on historical data and adapts to changing operating conditions, ensuring that the estimated measurements maintain sufficient precision for control purposes while providing the reliability advantage of not being subject to physical sensor failures.
Data Source
Figure 1~7
Figure 2
Figure 3
AI summary
A method is disclosed for controlling at least one compressor (13) in a gas compression system (60), the compressor being driven by an electric motor (3) powered by a drive (2). The method further comprises obtaining measurements of one or more process variables (21) for the compressor and/or compression system from sensors mounted in the compressor or the gas compression system and obtaining a value of at least one electrical parameter (31) from the drive (2) and/or the electric motor (3). Further an estimation (35) of at least one process variable is calculated and compared with a measurement of the process variable (q). The measurement is then either validated or else replaced with the estimated value (q est ) of the process variable. A computer program for carrying out the method and a compressor and compression systems employing the method are also disclosed.