Compressor Overconsumption Detection via Statistical Z-Score Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for tracking the performance of industrial compressors, such as air, oxygen, or nitrogen compressors, fail to accurately detect overconsumption due to changes in statistical behavior of input variables like temperature and pressure, leading to high false alarm rates and inadequate maintenance.
Innovation Solution
A method that calibrates a model of electrical power consumption during a learning phase and uses real-time measurements to calculate an overconsumption score based on a statistical test, incorporating a rolling time window and non-linear models like neural networks, to provide reliable overconsumption and invalidity indicators, and a compressor selection indicator.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a simple power consumption model is used for tracking, then the device complexity is reduced, but the measurement precision of overconsumption detection deteriorates due to false alarms from statistical behavior changes
Solution Approach 1:
The patent transforms the simple power ratio comparison into a statistical hypothesis test framework. It changes the parameters from raw power values to standardized statistical variables (Z-scores) that account for the statistical behavior of input variables. This allows the system to maintain relatively simple device complexity while dramatically improving measurement precision by filtering out false alarms caused by natural statistical variations in operating conditions.
2Reliability
If the detection threshold is increased to reduce false alarms, then the reliability of diagnoses improves, but the productivity of maintenance scheduling deteriorates due to missed detections
Solution Approach 1:
The patent implements a feedback mechanism where the statistical properties of input variables (temperature, pressure, flow rate) are continuously monitored and fed back into the detection algorithm. The system calculates the standard deviation and mean of these variables over time, and uses this feedback to dynamically adjust the detection threshold through statistical standardization. This ensures that the threshold adapts to changing operating conditions, maintaining high reliability without missing actual overconsumption events, thus preserving maintenance scheduling productivity.
3Measurement precision
If a three-month calibration period is used, then the model accuracy improves, but the time required for model updates increases
Solution Approach 1:
The patent transforms the static three-month calibration approach into a dynamic, adaptive system. Instead of requiring lengthy periodic recalibrations, the system continuously updates its statistical parameters (mean and standard deviation of input variables) in real-time as new data becomes available. This dynamic adaptation allows the model to maintain high accuracy by automatically adjusting to changing operating conditions, eliminating the need for lengthy fixed-duration calibration periods and reducing the time loss associated with model updates.
Data Source
AI summary
Method of tracking the performance of an industrial appliance (E), in which an estimate (We) of the quantity representing the electrical power consumed is calculated, in real-time, using the model (M), based on the values of the set (F) of operating parameters of the industrial appliance (E), the difference between the measured value (W) of the electrical power consumed (W) and the estimated value (We) provided by the model (M) is calculated, to obtain the overconsumption value, based on a statistical test on the overconsumption variable, using the distribution of this variable, an overconsumption score is deduced from this corresponding to a probability of overconsumption (p1) and an alarm is triggered if the probability of overconsumption (p1) exceeds a given overconsumption probability threshold (sp1), so constituting an overconsumption indicator.