In this post I’m going to give you an introduction to the Tyde-portal. For this purpose, I’m borrowing Glitre Energi’s Tyde deployment (they’re calling it aiKnow). In case you were wondering ; when you have your own Tyde deployment, like Glitre has, you are free to call it whatever you want to. We’ll even help you style it according to your preferences.
The first thing you will see when you enter the URL of your deployment is the log-in screen. Since in this case, Tyde is running on Microsoft Azure, the sign-in process is taken care of by Active Directory (AD). Signing-up is done from within the Azure portal.
Typically, you will just log in with the same credentials as you have for your work. It’s basically the same user as you use for logging into Office365, if you have that.
AD also has a nice 2-factor authentication mechanism. So if some part of your log-in attempt seems suspicious for AD, like an IP-address not yet seen, you will receive a SMS with a code to verify that you is you.
After you have successfully logged in, you will see the Tyde dashboard. It’s purpose is to give you some overview of your powerplant, with some of your favourite metrics shown. You’ll also see a map showing the locations of your powerplants. Notice that you can choose which metric you want to display here, by clicking at the various tiles. In the image you can see that I have clicked the right-most upper til, which brings up a dialog where i can search for- , and select a datasource/sensor.
On the left-side, you can see the main menu. The items in there can be expanded to give access to other pages of the portal.
Expanding the ‘Power Plant’ item in the menu, and the clicking ‘Asset Diagram’ will show you the image that you see here. The asset-diagram is an overview of the assets in your currently selected powerplant. Here you can see all assets placed into an hierarchical structure, which describes how they all relate to each other. This structure may be built in several different ways, for example by using the built in tools (notice the ‘+’ button in the lower-right asset, click it to create a new children asset).
We also display 2 types of flags in the asset-boxes. The down-pointing arrow means that the asset has a child-asset further down the hierarchy which requires some attention. The warning-sign indicates that this particular asset needs some attention.
The asset-diagram represents the context of the data in Tyde, as we will see next. Click any asset in the asset-diagram to proceed to the asset-view.
When you have clicked on an asset-box from the asset-diagram, you will see the asset-view. Starting from the top ; you’ll see some general information about the asset (name, vendor, IEC-81346 class etc.) and which groups of sensors are related to the asset. In the example-asset in the image, we see that there is only one sensor-group, but there may be more in other cases.
Further down, you will see an anomaly-calendar. The top level has a row of boxes representing months. A red color on the box, indicates that some anomalies has been found in that month, and a green color indicates that there are no anomalies. Clicking a month-box , updates the row of boxes below, which represents days. Day-boxes has the same meaning of the colors. Clicking on a day-box will show you the anomaly measures of this particular day.
Anomaly-measures are output values from a neural network model that has been trained with sensors and other data related to this particular asset. The input to the model are sensor-data from the same sources used during training of the model. Therefore, you will see that you get new anomaly-measures when new sensordata become available to Tyde. Anomaly-measures are values between 0 and 1, and associated with a timestamp (yes, they are timeseries, just like sensordata). 0 indicates perfectly normal, and 1 indicates perfectly non-normal. Sections with high anomaly values, as compared with their surroundings, are being marked with a red area. A red area means that Tyde thinks that this is an actual anomaly, happening in your asset. You can use the sensordata-view below the anomaly-measures to look at the sensor-values of the asset, at the time when the anomaly happened.
You may agree or disagree with the model. If you click on the red area representing an anomaly, a dialog will appear which isolates that particular anomaly so you can treat it individually. You can put a flag on it, indicating whether or not you agree with it being an actual anomaly. The algorithm which performs the training of the model will see those flags, and eventually learn to emphasize, or lower anomaly-values according to your flags. The flags are feedback to the models, on how they are doing, and they are learning from it.
Notice also that you can zoom and pan in the data-viewers.
Now, it could be that a particular anomaly is of particular interest for you, and you decide to investigate further. We got some tools to assist on that, described next
All aspects of Tyde has an own discussion-thread associated to it. The powerplants has their own thread, the assets has their own thread, the sensors has their own thread, and so on. For the particular asset we were looking at above, a Trash Rack, we discovered an anomaly and we been out to investigate further.
When clicking on the ‘thread-view’ button, as indicated in the image, you will see the thread-view. Here you can write some text, upload some images, videos or other files, and have a discussion with your colleagues. You can also use this functionality from the Tyde-app on your mobile-phone, making it easier to make images and videos and uploading them to the thread. (Note that the example in the image is made up, for the purpose of this explanation).
Going back to the asset-view. In the box which shows you the various sensorgroups that you have, you can click on a sensorgroup to go the sensorgroup-view. Here you see that there is a data-calendar similar to the anomaly-view. But in this case, the colors does not have any significant meaning. You basically click on them to browse the sensordata. Note that we display all sensordata from the sensors within the same graph-widget. It means they share the value-axis, and this may not make sense, depending on the engineering-units of the sensors. You can show/hide particular sensors by clicking their names, in the left-part of the data-view. Below the data-view, you can see a correlation table indicating correlations between the sensors in the group (don’t be fooled though ; “correlation does not imply causality”). And finally, you see sensor-boxes that you can click on to get to the single-sensor view, explained next.
The sensor-view will show you clickable aggregates (data-months, data-days) that you can use to inspect data at various levels of detail. For instance, the top-part shows month-boxes, and they serve as a long-time-trend view. Each box is built from statistics (mean, max-, min-values, standard-deviation). Clicking them will update the ”days” view, showing more details about each days in this particular month. Clicking the days, again, will just show the raw sensordata from that day.
The bottom part shows even more statistics from the timeseries produced by this particular sensor.
The sensor-view represents the bottom of the data-context. Furthermore we will discuss other aspects of Tyde
Going back to the far-left menu, below the asset-diagram, there is an item named ‘sensors’. Clicking in that will bring up a table containing all sensors know by Tyde. That can be a lot of course, but don’t worry. The table is indexed, so you can search for what you want. In the example in the image, i searched for “strøm” to limit the amount of matching sensors.
What you can do next is to click on the “eye” to the far-right to get to the sensor-view, which we discussed above. But you can also mark the sensors that you want, and then make a new sensor-group out of these sensors. Notice the “grid” button in the top bar, to the left.
If you click that button, a temporary group will be created, which you can store if you want to have easy access to it later.
Here you can see the sensorgroup that we just created. We covered the sensor-group view already, but you can notice that it does not have the correlation table yet. That is because your new sensor-group has not yet been discovered by the algorithm that calculates the correlations. It’s gonna appear shortly.
As I mentioned, currently the sensorgroup is a temporary construction. It will be gone, if you don’t save it. There is a save button in the top bar, to the left. Clicking it will save it with a default name (Local Group), from there you can edit the group to give it another name and change the sensor selection, if you wish. See the image below.
We covered a lot of Tyde-content in this post. But there is still a lot more to discuss. We will be covering these topics again in deeper detail and also other topics in future posts.
Thanks for reading 🙂