APRIL 202319THE ROAD TO DIGITIZING MANUFACTURING: DEMYSTIFYING IIOTCauses, Effects, Events, Data Compression, Data StorageIn general, there will be recording of two types of data: causes and effects. Production and downtime are examples of 'effects', and will usually be viewed against some other data i.e causes; for instance, which machine, which product, which operator, which shift, what reason for stoppage etc. Each change of cause is recorded as an 'event'. For example, the completion of a particular batch of raw material or the change of a specific part or operator suggests the cause has changed. Unraveling the relationship between cause and effect is the key to the IIoT system.Recording the data is at a different level of granularity from its usage. When data is recorded, it would be on a per second or millisecond basis generally. But when you look at the data, this level of granularity is not required. So the data can be compressed into per minute or hour for instance when viewing live data. Again if you are looking at historical data you may tend to look at it in days, so the data can be compressed into days. The storage of data needs to be in a manner that is consistent with its expected usage.DatabasesThe database needs to capture data in a manner that cause and effect can be correlated. Also, it should be possible to isolate and study parts of the database, like specific shops, machines, products, dates, and more. The data needs to be extracted and presented quickly and efficiently. An SQL data base should get the job done although other options may be available. MySQL is an open source system which enables interaction with the data base and would be my weapon of choice.Where this data is to be stored is another decision, at the 'edge' or in the cloud?The 'Edge' & The 'Cloud'There are several computations that should be done at the edge (or 'on premises') to avoid latency, particularly in time sensitive computations. It is also important where connectivity to the cloud is not good as it provides temporary storage as well. For instance you need to know when some parameter has crossed a limit, to take immediate corrective action, it is not optimal for the data to be sent to the cloud, be processed and come back as output. So, all real time activity should be at the edge, with only the historian outside- either on a computer (which can be on premise) or on the cloud. Data compression, which I talked about earlier, can also be performed at the edge.Having a computer on premise that does all the heavy lifting enables disconnection from the internet. And a second computer that mirrors the historian and with internet connectivity enables the best of both worlds.
< Page 9 | Page 11 >