SACA Factory Data Analytics 4 — Questions and Answers
Question 1: A manufacturer collects vibration data from 500 sensors every second. Which storage architecture best handles this volume?
- Relational database with row-based storage
- Time-series database optimized for high-frequency writes (Correct answer)
- Flat CSV files stored on a shared network drive
- Spreadsheet with auto-archiving macros
Correct answer: Time-series database optimized for high-frequency writes
Time-series databases like InfluxDB are architected specifically for high-frequency, timestamped data ingestion from many sources.
Question 2: What does a scatter plot between tool wear and surface roughness primarily reveal?
- The frequency distribution of roughness values
- The correlation or relationship between the two variables (Correct answer)
- The control limits for surface roughness
- The trend of tool wear over calendar weeks
Correct answer: The correlation or relationship between the two variables
A scatter plot visualizes the relationship or correlation between two continuous variables.
Question 3: In predictive maintenance, what is 'remaining useful life' (RUL)?
- The total operating hours a machine has logged since installation
- The estimated time until a component will require replacement or failure (Correct answer)
- The scheduled maintenance interval defined by the OEM
- The warranty period remaining on a piece of equipment
Correct answer: The estimated time until a component will require replacement or failure
RUL is the predicted time or operational cycles left before a component fails, enabling proactive maintenance scheduling.
Question 4: Which statistical test would you use to determine if defect rates differ significantly across three production lines?
- Paired t-test
- One-sample z-test
- Chi-square test or ANOVA (Correct answer)
- Pearson correlation coefficient
Correct answer: Chi-square test or ANOVA
Chi-square tests compare categorical outcomes across groups, and ANOVA compares means across three or more groups.
Question 5: What is feature engineering in the context of factory machine learning models?
- Installing new sensors to capture additional signals
- Creating derived input variables from raw data to improve model performance (Correct answer)
- Selecting the best ML algorithm for a given dataset
- Tuning hyperparameters of a trained model
Correct answer: Creating derived input variables from raw data to improve model performance
Feature engineering transforms raw sensor data into meaningful derived variables, such as rolling averages or rate-of-change, that better represent process behavior.
Question 6: A factory dashboard refreshes every 30 seconds with live OEE data. This is classified as:
- Batch analytics
- Near real-time analytics (Correct answer)
- Historical reporting
- Prescriptive analytics
Correct answer: Near real-time analytics
Near real-time analytics involves very short latency (seconds to minutes) between data collection and display, contrasted with true real-time (sub-second) or batch processing.
Question 7: Which concept describes the practice of training a predictive model on one production line and applying it to a new line with minimal retraining?
- Cross-validation
- Transfer learning (Correct answer)
- Ensemble modeling
- Hyperparameter optimization
Correct answer: Transfer learning
Transfer learning reuses knowledge from a pre-trained model and fine-tunes it on a new but related domain, reducing the data and compute needed.
SACA Factory Data Analytics 4
A manufacturer collects vibration data from 500 sensors every second.
Which storage architecture best handles this volume?