When a customer installs a SensorBot on a new machine, we follow this process to calibrate the SensorBot and get data streaming for the asset its connected to.
This process can also be followed for already installed SensorBots, you can skip certain steps (i.e. associating the SensorBot with an asset, etc).
NOTE: Some tools used in this article are being replaced with a new SensorBot management web app (Accumine Edge). If you see the below warning when going to Settings > Gateways or Settings > SensorBots in https://cloud.accuminetech.com , click Open Accumine Edge and use the new version but keep in mind that steps (2) and (3) in this article will be done differently (same overall concepts still apply though).
1. Get required info from customer
We need a few pieces of information:
- SensorBot Serial Number that's printed on the side (or on it's underside for older SensorBots)
- Inputs used and what the signals represent. We need at least one of these signals to calibrate a SensorBot:
- Signal that tells us when a part is produced
- Signal that tells us when the machine is running
- Asset that should be configured (i.e. what machine is the SensorBot hooked up to). It helps if they add the asset themselves in https://cloud.accuminetech.com (Settings > Assets) otherwise you'll create it
2. Associate the SensorBot with a Gateway and Asset
In https://cloud.accuminetech.com go to Settings > SensorBots. The SensorBot will be automatically added to the list of SensorBots once it is auto-discovered by the SensorBot and will look similar to this:
If you click on the row, a modal will appear to the right that looks like:
Select the asset its associated with in the Asset Commission dropdown and the closest gateway in the Gateway Selection dropdown. If you are unsure of which gateway is closest, ask the customer.
3. Configure the appropriate sampling
Keeping the right modal open (or re-opening if you closed it), click on the button that says Data Sampling Editor:
If there's a Signal that tells us when the machine is running, configure the Stream # that matches the Input # on the SensorBot like this:
This is telling the SensorBot to read Input 1 every 100 millseconds and calculate the average when you've read Input 1 50 times. Send this calculated value to the Gateway.
The value here will range between 0-1024 for SensorBots showing a version of 3.x.x and between 0-36,656 for SensorBots showing a version higher than 3.x.x. The lower threshold represents 0 volts and the higher threshold represents 5 volts (version 3.x.x) and 24 volts (all versions higher than 3.x.x).
Typically, this signal will oscillate around a certain voltage when the machine isn't running and at a different value when the machine is running. For example, let's say the signal represent current (amperage) drawn by the machine. Current will be low, even 0, when the machine is not running and higher when the machine is running.
If there's a Signal that tells us when a part is produced, configure the Stream # that matches the Input # on the SensorBot like this:
This is telling the SensorBot to read Input 1 every 100 milliseconds and count the number of times the signal went above 1,000. After 50 reads, send the total count to the Gateway.
The value here will be 0 or larger and tells us the number of rising edges (https://en.wikipedia.org/wiki/Signal_edge) that occurred within the 5 second period. The threshold of 1,000 tells the SensorBot at what value to calculate the rising edge, keeping in mind - 0 represents 0 volts and the 1023 represents 5 volts (version 3.x.x) and 36,565 represents 24 volts (all versions higher than 3.x.x).
For example, let's say the signal represent a proximity sensor that detects outgoing completed parts at the end of the line. The sensor will stay LOW (0 volts) when it does not detect a part and go HIGH (24 volts) when it detects a parts. In essence, we want to count how many times the sensor goes from LOW => HIGH as that tells us the number of parts that were completed.
Some sensors work in the opposite way - the sensor stays HIGH when it does not detect a part and goes LOW when it detects a part. The above sampling configuration is typically fine for this situation as well but if you encounter incorrect data, try switching the Type field from "Greater or equal" to "Less or equal".
When finished, click Save Changes and close the right modal.
4. Verify raw data
Find the SensorBot you're calibrating in the list and click on the Data Live Stream button at the far right of the row:
You'll see a new modal appear showing the live data from each stream:
At this stage you can engage the customer to help you verify that (1) they've connected the appropriate signal, (2) they've connected the appropriate signal correctly and (3) you've configured the sampling correctly.
If there's a Signal that tells us when the machine is running, you can ask the customer to let you know when the machine is running and when it is not. The objective here is to come up with a cutoff value that we can use to determine the status of the machine. For example, if the value in the Live Stream shows a range of 0-50 when the machine is not running and 500-1000 when it is running, a possible cutoff value is 200. When the value is below 200, the machine is not running and when the value is above 200, the machine is running. We will use this cutoff value for the next step - creating a data flow.
If there's a Signal that tells us when a part is produced, it is easier to finish calibration and then have the customer verify the counts in the real-time dashboards within https://cloud.accuminetech.com . Good options include Real-Time > Production Display.
As the customer becomes more familiar with the SensorBot and installs more and more, we often skip this verification stage as they're correctly installed on similar machines with similar equipment (i.e. the same proximity sensor make/model may be used across 30 machines).
5. Create data flow
Take a look at https://support.leading2lean.com/hc/en-us/articles/9165325536781 for a quick overview on data flows.
First, log into the customer's site at https://cloud.accuminetech.com and go to the Data Flows section (click on your username in the top right corner and select Data Flows). Select the associated asset in the dropdown menu and then add a new flow called "Machine Monitoring". A new page will appear with a window pop-up, select the Start From Template option.There's four common scenarios, pick the appropriate one based on the signals available and customer preferences.
Here's a video that walks you through the above steps: https://youtu.be/f2UQhs3r1vk
Scenario 1: If there's only a Signal that tells us when the machine is running and customer is not interested in part count
Step 1: Delete the existing input transform called "utilization" (blue tile) by clicking on it and pressing the backspace or delete key (note: if using a Mac, use fn + delete). Replace it with the "utilization" input from the "Inputs (Selected Asset)" section within the left sidebar and connect it to the Map transform (yellow tile) by clicking on the output square of the outgoing tile and drag your mouse to the input square of the incoming tile.
Step 2: Click on the Map transform (yellow tile) and replace the value on the 2nd line with the threshold you've estimated from the previous steps. When the utilization input is higher than this value, the asset will be considered to be in-cycle (running).
Step 3: Click the Save button:
Scenario 2: If there's only a Signal that tells us when the machine is running and customer wants us to use that signal to extrapolate part count
Follow the steps from Scenario 1
Scenario 3: If there's only a Signal that tells us when a part is produced
Step 1: Delete the existing input transform called "part_count" (blue tile) by clicking on it and pressing the backspace or delete key (note: if using a Mac, use fn + delete). Replace it with the "part_count" input from the "Inputs (Selected Asset)" section within the left sidebar and connect it to the Filter and Timeout transforms (yellow tiles) by clicking on the output square of the outgoing tile and drag your mouse to the input square of the incoming tile.
Step 2: Click the Save button:
Scenario 4: If there's a Signal that tells us when the machine is running and a Signal that tells us when a part is produced
Step 1: Delete the existing input transform called "utilization" (blue tile) by clicking on it and pressing the backspace or delete key (note: if using a Mac, use fn + delete). Replace it with the "utilization" input from the "Inputs (Selected Asset)" section within the left sidebar and connect it to the Map transform (yellow tile) by clicking on the output square of the outgoing tile and drag your mouse to the input square of the incoming tile.
Step 2: Delete the existing input transform called "part_count" (blue tile) by clicking on it and pressing the backspace or delete key (note: if using a Mac, use fn + delete). Replace it with the "part_count" input from the "Inputs (Selected Asset)" section within the left sidebar and connect it to the Filter transform (yellow tile) by clicking on the output square of the outgoing tile and drag your mouse to the input square of the incoming tile.
Step 3: Click on the Map transform (yellow tile) and replace the value on the 2nd line with the threshold you've estimated from the previous steps. When the utilization input is higher than this value, the asset will be considered to be in-cycle (running).
Step 4: Click the Save button:
6. Final data verification
Lastly, either yourself, the customer or both of you can verify the data by using Accumine Cloud. A few recommended views to do so:
Real Time > TV Display
Use to verify real-time status (i.e. running or not running)
Real-Time > Production Display
Use to verify part counts
Reports > Timeline
Use to verify running/not-running status over a historical timeframe
7. (Optional) Adjust the Timeout transform value
In every scenario, you may have noticed the Timeout transform (yellow tile) that has a value of 60,000 milliseconds. What this means:
Scenario 1, 2, 4: The asset will be in-cycle for 60 seconds since the last data point that met the in-cycle threshold.
Scenario 3: The asset will be in-cycle for 60 seconds since the last part produced.
This value is fine for most assets but can be adjusted if the customer would like the status to change slower or faster.