When we talk about "factory data," we often talk about Factory sensor data.
However, there are other data, and even if sensor data is used, it is often useful to Linking different source data. Analysis of Type 3 (Labeled Data) is an example.
MES is called a "manufacturing execution system," and some factories use it to manage which equipment to put into which equipment, when, and which lots.
Quality data may be included in the MES as the output of inspection equipment.
Quality data may be recorded in the database only for what is needed for shipment, and the rest may be stored on paper.
Factory sensor data is often not in the database.
If you are collecting data but not storing it anywhere, you need to create a mechanism to store it first. Even if it is not in the database, it may remain in the terminal used to run the machine.
Log data about how and when a machine moved can be useful.
Since it is not something that we usually see, it is more difficult to store in the database than sensor data, but it may be stored in the terminal used to operate the machine.
This data is valuable data, but since numerical values are mixed in the text data, preprocessing of the text data is required. Sometines it can be handled by the functions of EXCEL. However, machine data is always recorded with rules, so you only need to read the rules and program them, and there is no difficulty like natural language processing.
Mostly, it is written by a person and is often recorded on paper. There are various convenient terminals, but in terms of "anyone, easy, and cheap", paper is still effective.
When there was a pile of paper, I thought, "I can't do anything about it!" or "I can't analyze everything unless I enter it manually or digitize it with OCR!" It tends to be a story. However, even if the data is still on paper, it can lead to great results just by reading it diagonally and roughly classifying it. In addition, even if you digitize it, if you use only a small part, for example, only the part that matches the data you want to link, it may lead to speedy results.
The data you want to use to Problem solving and task achievement is often nowhere to be found. There are several reasons for this, but for example, unsolved problems are not in the form of "data" because the actual situation is not well understood.
In such a case, the data that you measure a little by yourself can be valuable data.
It's not something I'm proud of, but by measuring it as an author who is not familiar with the measurement work, I noticed something that I didn't notice with the data of an expert who usually measures (data measured by adhering to a standardized procedure), and it led to the solution of the problem.
For measurements that require specialized knowledge, it is good to ask someone who is knowledgeable to cooperate, but if you try to measure yourself as much as possible, It can help you better understand that data and help you solve problems.
If you have all the data, even if you don't know the theory of Cause and Effect Analysis, just graphing the data can solve the problem.
NEXT Factory sensor dataTweet