G5 Consulting & Engineering Services, David Schultz
| No comments yet
A dashboard provides a quick way to view the overall performance of your plant, whether it is operational or financial. Very quickly, you are able to drill into any area of concern. But their development is not as straight forward as people would like. Often times projects fail because a lot of time and resources are spent before anything meaningful is produced. One of the main reasons for this is due to lack of data or data in a non-useable format. As you begin or continue your industrial transformation and asset management journey, here are some thoughts on how to tackle the data challenges that lead to failed dashboards efforts in your organization.
Get Rid of the Clipboards
While clipboards are not as common as they once were, there continues to be a lot of data that is collected manually. Items like batch reports or shift reports often end up in a drawer, only to be looked at when there is a problem. The tragedy is that there is a lot of useful information on them. To get around that, companies will enter this data manually into a spreadsheet or other database, but this process is prone to errors. Moreover, the data may not be readily available to another system.
Replacing manual data collection methods with electronic devices like tablets is a great place to start. Not only is data collected in real time, the user interface can provide guidelines and rules for acceptable ranges. This gives immediate feedback to a potential problem as well as ensures data is entered correctly. Of course, you will want to make sure this data is readily available to other systems. More on this later.
Add Some Sensors
One of the features of the Industrial Internet of Things (IIoT) is the ability to add and connect sensors to your current systems rather inexpensively. This includes process instrumentation for additional process variables, like temperature, pressure and level. It also includes asset condition management data like vibration, infrared and ultrasound. Again, rather than collecting this data manually, you want to capture all of the new sensor data in real time.
To make it more simple and less expensive many of these sensors are available in a wireless system. But before you start adding sensors everywhere, you will want to consider wireless security. There are a number of wireless protocols that are used for home systems, but are not appropriate for industrial use. Unfortunately, many newcomers to the market are using communication technology geared for home use. Be sure to understand the differences and the risks.
The Data Store
Now that you are collecting data electronically and with additional sensors, you need a place to store the data for future use. When selecting a data storage technology, it is important to remember the four Vs of data. These are volume (how much data), variety (what types of data), velocity (how fast do the values change) and veracity (is it good data). The good news for a dashboard is that most of the data discussed so far is collected in a time-series format. That means it is a value at a point in time. Because the new data points are not part of the safe operation of the plant, it is likely the volume is low, only one type of it, it doesn’t change quickly and generally good quality (assuming you didn’t use a clipboard).
While they historically have been used for process control, modern SCADA systems are great for data storage. Many come with a built-in historian that will record this data, and uses logic engines to minimize storage space. While technically it is still a database, the querying and retrieval of data is much more streamlined, as data is summarized for longer time spans. For these new data sources, I discourage the use of conventional process historians, as the cost for tags and points is not trivial.
Garbage In. Garbage Out.
Before we move on, I want to add a quick note on veracity. Industrial transformations are built on the premise that more data will lead to better decision making. It is critical to ensure the data you are collecting quality data. While the data is likely good quality, you should consider the implementation of a calibration program for new measurements that are deemed critical to the effort. Along with sensor technology, there is a lot of new calibration technology available with a much lower investment. Keep in mind, your dashboards will be used to make decision and you don’t want to make decisions based on bad data.
The Main Event
It is quite common for the creator of a dashboard to develop queries that define the conditions of the requested data. This leads to several problems. First, it consumes a lot of time as this process will be done for every element on the dashboard. Second, it requires a significant amount of processor power on the database server to provide a result. Finally, it uses significant network resources as the query payload is sizable. While this may not be a problem for a few dashboards, consider how it will affect your system when many dashboards are rolled out across your organization.
To solve this problem, you need to create and use an event database, which performs two critical tasks. The first is the detection of events, like changes in shift, batch and downtime. While these event may all exist in disparate systems, the event database will detect the change and record the event. The second is the application of a stored procedure. Rather than using a convoluted query (described above) the procedure will run when the event occurs and store the outputs in a database.
At this point we have created a fantastic infrastructure. You have automated you data collecting process and taken steps to ensure its quality. It is being stored in a system that makes it easily retrievable by other system. You have summarized time-series and event data. However, you are not quite ready for your dashboard. Skipping the next step is what has commonly lead to the failure of a dashboard effort.
A concept that has emerged as part of a transformation effort is a unified namespace. Again this is critical to your dashboard effort. What this allows you to do is create a common data source with a consistent naming convention throughout your enterprise. More importantly, every system now becomes a node with the namespace acting as the central data repository. Along with the aforementioned newly created data, all systems will take advantage of this namespace.
You can think of the namespace structure in terms of a plant model. The enterprise will be at the top, followed by the plant, followed by unit or line, asset, and so on. Names can include PLC tag values, or MES batch numbers or warehouse inventory levels. All of this data can be quickly accessed from an analytics or visualization system. To be sure, there is much more to this concept, but for the purposes of a dashboard effort, you will definitely want to explore it.
Paradise by the Dashboard (Light)
Congratulations! You are now ready to create a dashboard. The data is now high quality (human and sensor errors are reduced) and available in real time. Events are already stored and summarized in the event database which make for a simple query. It uses a simple naming convention that enables easy retrieval, with most, if not all, of the business systems connected.
Along with a dashboard, the same system can be used for analytic tools. PowerBI,Seeq, Splunk and Tableau will all take advantage of readily available data to build ad-hoc queries and reports quickly. The consumers of the data will appreciate your efforts. The technology people will appreciate them, too. And you will likely get many accolades for delivering a dashboard effort that is on time and under budget.