Mechanical State Diagnosis Using Streaming Statistical Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information processing apparatuses for diagnosing mechanical apparatuses require high calculation capability and large storage capacity, which is not feasible for many microcontrollers and server setups due to memory limitations and management costs.
Innovation Solution
An information processing apparatus that includes a sensor data acquisition unit, internal variable holding unit, internal variable calculation unit, feature calculation unit, and state diagnosis unit, which processes sensor data by calculating internal variables and features without needing high-capacity computers, using a method that reduces the amount of data stored and processed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data is held for a set period to enable anomaly detection, then diagnostic accuracy is improved, but storage capacity requirements increase enormously
Solution Approach 1:
The patent extracts only the essential features from the raw sensor data by calculating statistical quantities (mean, variance, skewness, kurtosis) and holding these calculated features instead of the original high-volume sensor data. This extraction approach maintains diagnostic accuracy while dramatically reducing storage requirements from terabytes to manageable sizes.
Solution Approach 2:
The patent transforms the sensor data by changing its parameters - converting raw measurement values into statistical parameters (mean, variance, skewness, kurtosis). This parameter transformation reduces the data volume while preserving the diagnostic information needed to detect mechanical anomalies.
2Measurement precision
If high sampling frequency is used to capture detailed mechanical states, then measurement precision is improved, but data processing burden increases
Solution Approach 1:
The patent extracts essential diagnostic information by calculating statistical features from the high-frequency sensor data. Instead of processing all raw data points, it extracts mean, variance, skewness, and kurtosis values, significantly reducing the processing burden while maintaining measurement precision for anomaly detection.
Solution Approach 2:
The patent applies partial action by calculating only the necessary statistical features (mean, variance, skewness, kurtosis) rather than performing exhaustive analysis on all sensor data. This selective processing maintains diagnostic accuracy while reducing the overall data processing burden.
3Reliability
If all sensor data is retained for offline processing, then diagnostic completeness is improved, but memory requirements become prohibitive for microcontrollers
Solution Approach 1:
The patent extracts essential diagnostic information by holding calculated statistical features (mean, variance, skewness, kurtosis) instead of retaining all sensor data. This extraction maintains diagnostic completeness while reducing memory requirements to levels suitable for microcontrollers with limited memory capacity.
Solution Approach 2:
The patent segments the diagnostic process into two stages: online calculation of statistical features from sensor data, and offline analysis of these features. This segmentation allows microcontrollers to handle only the lightweight feature calculation in real-time, while more intensive analysis is performed later on the extracted features.
Data Source
AI summary
In an information processing apparatus, an internal variable holding unit holds a smaller number of values of an internal variable than N, the internal variable being sequentially calculated in time series based on sensor data acquired by a sensor data acquisition unit. An internal variable calculation unit calculates the internal variable corresponding to a (j+1)-th time point (j is an integer of one to N−1) based on the sensor data at the (j+1)-th time point and the internal variable corresponding to a j-th time point. A feature calculation unit calculates a feature by extracting a statistical characteristic included in the sensor data from the first time point to the N-th time point based on the internal variable at the N-th time point. A state diagnosis unit makes a diagnosis of a state of the mechanical apparatus based on the feature.