Time-Series Data Correction Around Equipment Stop Periods
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data processing systems inaccurately detect equipment anomalies due to variations in detection values caused by maintenance or sensor deviations, leading to incorrect machine learning models and thresholds.
Innovation Solution
A data processing apparatus and method that corrects data variations by identifying stop periods during maintenance, calculates the amount of variation in process data before and after these periods, and uses corrected data to generate a machine learning model for anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If data is collected continuously including during maintenance periods, then data quantity increases, but measurement precision deteriorates due to sensor deviations and maintenance variations
Solution Approach 1:
The patent segments the time-series data into multiple periods based on maintenance schedules, identifying which periods correspond to maintenance activities. By separating maintenance periods from operational periods, the system can selectively exclude or separately analyze data collected during maintenance, thereby preventing contaminated data from degrading the overall measurement precision while preserving the quantity of useful operational data.
Solution Approach 2:
The patent performs preliminary identification of maintenance periods before conducting anomaly detection or machine learning model generation. By pre-marking which time periods correspond to maintenance activities using maintenance management system data, the system prepares the data in advance to exclude or separately handle maintenance-period data, ensuring that only high-quality operational data is used for critical analysis tasks.
2Productivity
If machine learning models are trained on all available data including maintenance periods, then model development speed increases, but model reliability deteriorates due to data variations from maintenance
Solution Approach 1:
The patent extracts and removes data corresponding to maintenance periods from the dataset used for machine learning model training. By identifying maintenance periods through comparison with maintenance management system data and excluding these segments, the system ensures that training data contains only operational data, thereby maintaining high model reliability while still utilizing sufficient data quantity for effective model development.
Solution Approach 2:
The patent performs preliminary filtering of training data by identifying and removing maintenance-period segments before initiating machine learning model training. This pre-processing step ensures that the model is trained exclusively on clean operational data, preventing maintenance-related variations from compromising model reliability, while maintaining efficient model development through automated data preparation.
3Device complexity
If anomaly detection thresholds are set using all data including maintenance periods, then threshold calculation becomes simpler, but detection accuracy deteriorates due to data variations
Solution Approach 1:
The patent segments the data used for threshold calculation by identifying and separating maintenance periods from operational periods. By calculating anomaly detection thresholds using only operational data (excluding maintenance periods), the system maintains detection precision while keeping the calculation process relatively simple through automated period identification and data filtering based on maintenance schedule information.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
(SOLUTION) Provided is a data processing apparatus including: a data acquisition unit for acquiring time-series data regarding running of an apparatus; a stop acquisition unit for acquiring a stop of the running of the apparatus; and a correction unit for correcting, after changing a reference value according to an amount of variation in the data between before and after a stop period of the apparatus, the data using the reference value. Provided is a data processing apparatus including: a data acquisition unit for acquiring time-series data regarding running of an apparatus; a stop acquisition unit for acquiring a stop of the running of the apparatus; an output unit for outputting an amount of variation in the data between before and after a stop period of the apparatus; and a correction unit for correcting, after changing a reference value upon receiving an instruction corresponding to the output of the variation amount of the data, the data using the reference value.