This is what my table looks like

which is the best approach to design this database?

And I use postgresql in Hasura(Graphql engine that generate graphql server by define database with no code)to create query and mutation.

I try to pack session data(many rows of data like a point survey follow by point name that save in template) into array of JSON in in Observation_data column.
I only need to know which is the best approach to design this database.
I want to implement them to create graphql api for robot ui which use to query and mutation the inspection data of thickness of iron pipe.

If you have another method that better than that two choice you can suggest me.

I concern about ACID rule of RDBMS.