Before you begin
Asset Risk Predictor must be purchased separately from a Fiix representative. To learn more, see the Eligibility & purchasing section below.
This article provides answers to the following frequently asked questions:
What makes asset risk predictor different than other predictive maintenance or AI solutions?
Asset Risk Predictor is deeply integrated to OT layer and uses a multi-variable model, which includes inputs from multiple sensors, machine status, and recipe information to provide a 360 view into a machine’s health. It also offers fast time to value, as it can be up and running in as little as 2 weeks.
How is Asset Risk Predictor different than condition-based monitoring?
The main differences are:
- Automated learning: Unlike traditional condition-based monitoring where thresholds have to be manually set based on expert knowledge and data analysis, Asset Risk Predictor uses machine learning algorithms to learn what is normal for each asset automatically, eliminating the need to manually set up thresholds.
- Intelligent anomaly detection: While traditional condition-based monitoring might trigger false alarms due to small deviations or spikes in sensor readings, the ARP algorithm is intelligent enough to ignore these minor anomalies that are not indicative of a problem.
- Complex problem detection: Traditional condition-based monitoring often relies on simple thresholds that may not be sufficient to detect complex problems that involve a gradual change in condition over time or correlations between multiple data points. Asset Risk Predictor, however, can identify these complex patterns, allowing for earlier detection of potential issues.
- Predictive maintenance: Unlike traditional condition-based monitoring, which tends to focus on current conditions, ARP includes next-day prediction capabilities. This feature facilitates predictive maintenance, allowing potential issues to be addressed before they cause a failure.
- Reliability metrics and predictions: Given failure codes from PLCs, Asset Risk Predictor is capable of calculating reliability metrics, which are key indicators of an asset's probability of performing its intended function without failure over a specified period of time. By analyzing these metrics, Asset Risk Predictor can predict not only the risk of failure but also other aspects of asset performance, such as the expected remaining useful life of the asset.
What is the difference between using a threshold-based system and the Asset Risk Predictor?
Threshold based systems are often used to indicate when a parameter or metric is outside the bounds of their specs. When you go out of spec, you get an alarm. Often these systems are single-dimensional and don't give much warning before failure.
Asset Risk Predictor learns the signature of how the sensors behave during normal operation, and also throughout the year. It also is multi-dimensional, so it can indicate risk based on small movements when it finds patterns of more than one sensor movement outside of “normal”. It can also give warning prior to failure, allowing you to schedule maintenance before actual failure happens and in a controlled and planned manner.
How do we know the data and predictions we’re receiving are accurate?
Technicians can verify the accuracy by looking at the raw sensor values and operating statuses during the risk events we specify. If you provide failure logs, accuracy can also be measured automatically.
How are false positives handled?
For now, we need to review the machine to understand if there is a false positive. More often than not, the sensors are in fact drifting, maybe not because the machine itself is failing, but because an upstream process could be causing changes in the behavior of the machine that we are monitoring. If this is not the case and there is in fact a poor model causing false positives, we would trigger a retrain or wait for the monthly relearn to self-correct.
Eligibility & purchasing
How do I purchase Asset Risk Predictor?
Contact your Customer Success Representative or reach out to our product experts at firstname.lastname@example.org.
Who is eligible for Asset Risk Predictor?
Anyone can purchase Asset Risk Predictor, even if you're not already a Fiix customer.
For Fiix customers who purchase Asset Risk Predictor, it will be enabled in your existing CMMS. To learn more, see Overview: Asset Risk Predictor setup.
For non-Fiix customers, we'll create a Fiix CMMS for you where you can access Asset Risk Predictor. You will receive one Enterprise license with your purchase. If you need additional licenses, you can contact your Fiix representative to purchase them separately.
Are there any disqualification criteria that could prevent us from using Asset Risk Predictor?
If there is no statistical “normal” pattern to how the asset and sensors report data, this can cause a very wide range of “low” risk points. The machine should have a normal pattern to its data, and working towards a reduced standard deviation on the sensor data can help with a good model.
If the asset has drastic swings in mean and the standard deviation is high or sporadic, it reduces the accuracy of predictions. We use very advanced algorithms and techniques to power Asset Risk Predictor, but you can review the Cp and Cpk calculations in the six-sigma methodology to gain a basic understanding of centering mean and how a high standard deviation negatively influences a process.
Further, assets like vehicles are not ideal, as Asset Risk Predictor should be monitoring an automated process like a manufacturing line, work cell, fabrication cell, or even calculated values that are derived from a process. For example, we could be calculating OEE in a MES and that value is updated in real time. A user could query that OEE value every second, five seconds, etc., and then feed data that into Asset Risk Predictor as a sensor value. Then you would get a high-risk score on that OEE “sensor” when OEE changes from its normal pattern of fluctuation to its new pattern.
Does the algorithm continue to learn what is normal past the initial training period?
Yes, Asset Risk Predictor uses the first 7 days as the initial training period, and then automatically retrains after one month. After that, it retrains every six months automatically.
How does the algorithm learn what is normal? Do we have to choose a “normal” week for implementation?
Ideally, we choose a week when the machine is behaving as we would like it to. However, we can retrain at any time if the machine does not run as expected during the training week.
What happens if something goes wrong with the asset during the training period?
Ideally the machine is sending out faults and Asset Risk Predictor will disregard these sections of data and wait for clean data to train on. If this is not the case and machine fault information is not available, then we would trigger a relearn.
Does the algorithm account for degraded performance over the time we use Asset Risk Predictor?
We’ve developed the retraining schedule for our algorithm with this in mind. By automatically retraining every six months, rather than on a more frequent schedule, we’re able to see degraded performance over time.
What types of assets are supported?
You can use Asset Risk Predictor with any assets that can be monitored using sensors.
Some examples include (but are not limited to):
- Industrial machinery, such as pumps, motors, and gearboxes, to monitor for abnormal vibrations that may indicate a mechanical failure or unbalance.
- HVAC systems, to monitor temperature, humidity, and air flow.
- Electrical systems, to monitor current, voltage, and power usage.
- Motors and generators, to monitor speed (RPM) and power output and consumption.
- Manufacturing equipment, to monitor temperature and humidity levels in the production environment.
- Automotive, to monitor temperature, humidity, speed and vibration, of engine, transmission and wheels.
- Wind turbines and other renewable energy systems, to monitor for abnormal vibrations, temperature changes and power output.
- Robotics, to monitor temperature, humidity, vibrations, speed, current and voltage for safety and performance optimization.
Related: What types of sensors can I use?
How does this solution scale? For example, could we use it for 1000 assets?
Yes, because it is a scaled cloud solution and it can technically be enabled for 1000 assets. If you want to enable Asset Risk Predictor for a large number of assets, contact your Fiix representative to discuss your options.
Is non-continuous manufacturing supported?
Yes, batch manufacturing is supported. This process relies on the machine having an indicator of when the machine is running. Asset Risk Predictor uses an IsRunning variable to let it know which data points occurred when the machine was online or offline. We also recommend using the Fault variable in the payload, which allows us to understand the running condition of the machine when the operator has it enabled. For example, maybe the operator is running the machine and it is in a slow cycle state or an overloaded state. Asset Risk Predictor considers this and adjusts how it interprets the data and risk score.
Are operations with different load supported?
Yes, we support operations with different loads. Any recipe information we receive is used alongside sensor data in the training model. This ensures that prediction results include the recipe details, allowing the algorithm to accommodate different loads without mislabeling them as anomalies.
Can I add or change assets later?
Yes, you can add or change the assets (and sensors) included in Asset Risk Predictor. The process for adding or changing assets and sensors is similar to the initial setup. To learn more, see Overview: Asset Risk Predictor setup.
There is a cost associated with adding or changing assets and sensors.
To add or change the assets in Asset Risk Predictor, contact your Fiix representative.
Do I have to pay to include recipes, fault codes, and machine statuses in my data stream?
No, you only pay for sensors/tags. Recipes, fault codes, and machine statuses are included for free.
The following is an example of a payload that is considered one sensor input:
“message”: “V6 Public API”,
“recipe”: “Recipe Test new “,
What if I have rotating assets that get an inconsistent amount of use? Will this impact the results in Asset Risk Predictor?
Not always. Asset Risk Predictor waits for the “IsOnline” flag to be true before using the data, so if an asset is offline, it should not be sending data or this flag should be false.
If we are swapping components on a machine, we are monitoring with Asset Risk Predictor and this action is causing false high-risk events, there are a few options. The easiest way is to incorporate this machine setup into the recipe field in the data package, and Asset Risk Predictor develops a custom model for this specific setup. For example, if the recipe looks like “ModelXYZ” we can adjust this to “ModelXYZ-Setup1”, “ModelXYZ-Setup2”, etc. Now, Asset Risk Predictor detects a change in the recipe based on the rotating assets or setups, and creates a custom model for each one.
Are there any types of assets that aren’t a good fit for Asset Risk Predictor?
Low-run assets are a bad fit, as we need (more or less) continuous data. To keep the system running optimally, we need at least 2400 minutes of data over 7 days. Therefore, if the machine does not run for 2400 minutes a week where the “IsOnline” flag is not true, and there are no faults, the system will begin to skip data points. These 2400 data points can be in a row or broken up, but must occur in a 7-day running span.
What types of sensors can I use?
Asset Risk predictor is designed to be sensor-type agnostic, meaning that it can work with all types of sensors.
Some examples include (but are not limited to):
- Speed (RPM)
- Current or power
What if I don't have sensors?
Fiix can introduce you to our network of partners who can support with installing sensors on additional equipment. Contact your Fiix representative for more information.
How many sensors are needed?
It depends on the type of asset. For a small asset, 3 sensors might be enough to detect anomalies. Typically, using more sensors leads to higher the accuracy and allows for better root cause analysis.
Can I have one asset with many sensors? For example, could I have 500 sensors on one asset?
While this is technically possible, having this many sensors on one asset makes it difficult to troubleshoot risks. Instead, we recommend splitting the large assets into "sub-assets" that correspond to different parts of the machine with their own sensors.
For example, instead of just having an extrusion line, you could create the sub-assets for the main extruder, feed hopper, cooling system, heating elements, cutting mechanism, control panels, and other relevant equipment.
How is it possible that the algorithm works with any type of sensor?
The versatility of Asset Risk Predictor comes from its fundamental operating principle. Our system works on the concept of “normal behavior”, which is learned from repeated observations. Just as our understanding of “normal” temperature changes throughout the year—expecting colder temperatures in winter and warmer in summer—our system also understands and adapts to these patterns.
In essence, any sensor that provides data consistently creates a pattern. It doesn’t matter what kind of sensor it is, so long as it provides a regular data stream. The machine learning model is not understanding the physical phenomena the sensor is capturing; it’s analyzing the behavior pattern of the data.
How are risks identified?
To identify risks, our smart AI algorithm first establishes a baseline for each asset you've included in Asset Risk Predictor. Then, it uses aggregated sensor data to determine the risk score for each asset and sensor. This score is generated by averaging the results of 3 different calculation methods, resulting in a more accurate score. It also considers other factors, such as the day of the week, so that it can identify patterns in your asset's operations.
Then, it uses this risk score to assign a risk level. To learn more (including how to change the risk level thresholds), see Change risk level definitions for Asset Risk Predictor.
Can I customize the risk thresholds?
Yes, you can change the definitions (i.e. thresholds) for risk levels. For example, you could change the definition of high risk to start at a risk score of 600 (instead of the default of 800). To learn more, see Change risk level definitions for Asset Risk Predictor.
What does the predicted risk level mean?
In the Asset Risk Predictor dashboard, we provide a predicted risk level. This is the risk level we predict the asset will be at tomorrow if no action is taken (such as maintenance or operational process changes) between now and the next report.
Can I receive alerts about risks?
Yes, if you have user or creator permissions for analytics, you can set up automated alerts in the Asset Risk Predictor dashboards. To learn more, see the Set up alerts section in Asset Risk Predictor tips & tricks.
How do I action risks?
You can action risks by:
- Monitoring Asset Risk Predictor daily and inspecting an asset whenever you see a medium risk.
- Enabling the automatic creation of risk-based work orders. To learn more, see Enable risk-based work orders.
Is there a way to automatically create work orders based on risk?
Yes, you can configure the CMMS to automatically create work orders in the following situations:
- When there’s a high risk score for any hour in the past day
- When there was a medium risk score for 3 consecutive hours or 6 hours total in the past day
Note: This only applies to assets that don’t already have open work orders. If an asset already has open work orders, new ones will not be created automatically, even when the above conditions are met. Instead, updated risk information will be added to the existing work order.
To learn more, see Enable risk-based work orders.
Will the CMMS keep creating work orders if I don't close the first one?
No, it doesn't create new risk-based work orders for assets that already have an open work order. Instead, it adds the latest risk information to the existing work order.
Data & security
What work order information is used for Asset Risk Predictor?
Asset Risk Predictor doesn't use work order data; instead, its models are based on data payloads coming from the OT system/sensors.
What physically enables this service?
Enablement is completed by submitting data from a data source to the API endpoint. Our model is based on connecting an edge data collector, aggregating the data by the minute, and submitting every hour. However, there is no reason why data from an existing historian can't be connected as well using an integration service (like Boomi, SSIS or any other ETL system), so long as we get data submissions by the minute, submitted only once per hour. If you are interested in creating your own connector, this is an option as well, as a simple HTTP client in the language of your choice can submit data to ARP.
How is sensor data sent to Asset Risk Predictor?
As mentioned in the previous answer, any system that can generate a JSON payload containing aggregated data of values can be used to send data to ARP. Data contained in a historian, db, data lake, and of course live data from OT sensors and PLC inputs can be used. For example, if you are collecting OEE and have that logged to a historian with live data, this value can be sent to Asset Risk Predictor, and if the signature changes from the trained model we will see risk for that “sensor” input rise. We do not need to think about sensors as just inputs from a machine; they can be results from a calculation on a process as well. Any numeric value can be used as long as it can be read by the minute, has a "normal" pattern to it, and requires attention when it deviates from "normal".
Is my data secure when sent to the cloud?
Your data is secure during transmission and at rest. Data can only be sent with your valid token and is encrypted during transmission. We are HIPAA, PCI DSS, SOC 1, and SOC 2 Type 2 compliant for data governance.
Why is some of my risk data blank?
We filter sensor data based on whether or not the asset is running. If your hourly or daily charts are missing columns, it generally means that the asset was not running at that time.
If the entire dashboard is empty, it means that the automated algorithm training has not been completed yet. This process takes 7 days, and requires at least 2400 data points.
If 7 days have passed, but the dashboard is still empty, it means that there isn't enough sensor data yet to complete the training. Once there is enough data, the dashboard will be populated automatically when training is complete.
Why can't I open the Asset Risk Predictor dashboard?
If Asset Risk Predictor has been enabled for your organization, but you can't access it, it likely hasn't been enabled correctly for your user account:
|If...||it means that...|
|Asset Risk Predictor doesn't show up in the feature menu at all||You haven't been assigned the necessary menu permissions|
|Asset Risk Predictor shows up in the feature menu, but you can't access either of the dashboards||You haven't been assigned an analytics seat|
|You can access the Asset Risk Prediction Overview dashboard, but you can't click an asset to open the Asset Risk Predictor dashboard||You've been assigned a viewer seat for analytics instead of a user or creator seat.|
Contact your administrator to update your user account settings. To learn more, see Enable access to Asset Risk Predictor.
What if I run out of users?
Your plan comes with a set number of users and analytics seats. If you've reached your user or analytics seat limit, you can do one of the following:
- Schedule dashboard delivery via email. If a user only needs to view the dashboard, you can set up Asset Risk Predictor to send it to them via email automatically. Although they won't be able to log in, they'll still be able to view the risk data via email. To learn more, see "Schedule dashboard delivery" in Asset Risk Predictor tips & tricks.
- Contact your Fiix representative to purchase additional licenses and/or analytics seats for users who need to be able to access the dashboards within the CMMS.