Substrate Processing Time-Series Segmentation for Abnormality Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for analyzing time series data in substrate processing apparatuses fail to accurately discriminate between normal and abnormal substrate processing, leading to incorrect assessments of processing quality.
Innovation Solution
A data processing method that identifies rising, stable, and falling periods in time series data and calculates evaluation values for these periods, allowing for accurate discrimination of normal processing by comparing these values with reference data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional time series data analysis methods are used, then the analysis process is simple, but the accuracy of discriminating normal and abnormal substrate processing is insufficient
Solution Approach 1:
The time series data is segmented into three distinct periods: rising period (from initial level to target level), stable period (maintaining target level), and falling period (from target level to initial level). By dividing the continuous data into these discrete segments, the system can apply different evaluation criteria to each period, thereby improving the accuracy of substrate processing discrimination without overwhelming complexity
Solution Approach 2:
The patent changes the evaluation parameters by calculating specific evaluation values for each period: rising time (duration of rising period), stable time (duration of stable period), and falling time (duration of falling period). These parameter changes transform the continuous time series data into discrete measurable metrics that can be accurately compared against reference values to determine processing normality
2Measurement precision
If detailed period-specific evaluation is implemented, then the discrimination accuracy improves, but the data processing complexity increases
Solution Approach 1:
The data analysis is segmented into three clear phases with distinct evaluation metrics for each phase. This segmentation simplifies the detection and measurement process by providing a structured framework where each period has its own evaluation criteria, making the overall complex task more manageable and systematic
Solution Approach 2:
The patent transforms complex time series analysis into simpler parameter measurements: rising time, stable time, and falling time. By changing the evaluation parameters from continuous waveform analysis to discrete time duration measurements, the system improves accuracy while reducing the difficulty of detection and measurement
Data Source
AI summary
A data processing method includes a period setting step of obtaining a rising period, a stable period, and a falling period with respect to time series data obtained in a substrate processing apparatus, an evaluation value calculation step of obtaining an evaluation value in the rising period, an evaluation value in the stable period, and an evaluation value in the falling period as an evaluation value of the time series data. In the period setting step, a period from when a control signal changes until the time series data falls within a first range including a target level is obtained as the rising period, a period from when the control signal changes until the time series data falls within a second range including an initial level is obtained as the falling period, and a period between the rising period and the falling period is obtained as the stable period.