Data collected by temperature sensors, brightness sensors, etc. is called sensor data.
In particular, the "sensor data" handled on this page is It is a fixed-point observation and continuous measurement at a period such as every 1 second or every 1 minute. In the field of environmental studies, it is used to constantly monitor places that are difficult for people to enter, and in the field of quality studies, it is used to manage factories. The data of the power monitor used for energy conservation in buildings is also sensor data.
Analysis of sensor data is a typical example of Time Series Analysis.
30 days is 43220 minutes (30 days * 24 hours * 60 minutes), so 1 days of data for every 30 minute is 43220 rows. If you try to handle this amount on a general computer, depending on the processing, it will not be possible to say "instantly". This is the situation with one-minute data, so if you try to get more detailed data to see instantaneous changes, it will be even more difficult.
Also, when analyzing sensor data, there are times when you want to pay attention to the relationship of many sensors. The data for the above number of rows increases by the number of sensors.
"Big data" is in the spotlight in the world. Sensor data is also big data.
Sensor data in factories looks quite different depending on the production method, and the approach is also different.
The quasi-periodic type is the temperature data of the oven when bread is taken in and out of the oven and baked.
In factories, it is a production method called the batch type. This is true not only for factories, but also for phenomena that events are repeating. There is no vibration data or accurate period like electrical signals.
Spectrum Analysis cannot be used to analyze exact periods. The key to analysis is to quantify the variation in parts of the Quasi-periodic Data Analysis that are not accurate periodic.
The flow type is a tunnel-type oven, which is the temperature of the oven when baking while passing through the oven. In factories, it is a production method called a continuous type.
Analyzing the relationship between weather data and events is also a flow type.
We sometimes approach the flow type with Condition Analysis and consider Time Difference of Cause-Effect, but it is quite difficult. An approach similar to Quasi-periodic Data Analysis is Reverse Time Aggregation.
Discrete is a production method that is somewhere between batch and continuous. For example, it is a method in which one product can be produced for each revolution that rotates mechanically and regularly.
Basically, it moves mechanically and regularly, so the period between them is quite precise and cyclical. Time variations, such as quasi-periodic types, result in data that happens from time to time.
Therefore, the discrete method differs from the batch method in that it selects only periodic times and allows the periodic data analysis method to be used for that range. You can investigate the differences in cyclical characteristics depending on the time of year. However, even if it is discrete, analyzing quasi-periodic data often tells you what you want to know.
NEXT Quality Engineering