I’m thrilled to announce the release of our latest feature, ExplainIt for Live Predictions, empowering factories to understand, explain, and trust real-time AI predictions.
Fero Labs ExplainIt for Live Predictions provides real-time factor-based explanations about production anomalies, giving AI-augmented factories greater production insight and confidence in live production optimizations.
Why we’re excited about ExplainIt:
- Provides real-time factor-based explanations into live production anomalies
- Aids in diagnosing or preventing production issues faster
- Jump starts a non-linear approach to root cause analysis to avoid biases and increase accuracy
It’s hard to trust what you don’t understand. ExplainIt builds trust by expanding live production knowledge.
This ground-breaking feature builds on our platform’s quest to provide transparency and explainability to AI-augmented factories so their live production knowledge can expand in realtime.
Humans will naturally distrust AI predictions. This is why we were first to market to develop our legacy white-box ML, which provides context and transparency into the models and algorithms that we use to make predictions and recommendations.
It’s also why we provide confidence levels on our predictions in simulated environments and live deployments.
ExplainIt expands on this same concept by disclosing real-time text explanations about abnormal conditions that our system observes within a current production. This empowers engineers to make more informed decisions and diagnose or prevent production issues faster.
Oftentimes, plants don’t discover issues like this until it is too late. Materials are wasted. Hours or days are spent after the fact trying to identify the root cause. Using our ExplainIt feature, plants can identify problems in seconds.
We consider ExplainIt as being similar to showing your work on a math test. Understanding the details behind the prediction builds trust for the engineer. The benefit of this is that engineers can learn about their production at much greater speeds than ever before, making them more informed and agile to react to their changing production conditions. They can also train others in what they've learned.
ExplainIt’s real-time text explanations include:
- Which subset of factors are unusual
- How each factor is unusual
- Degree of unusualness
- Typical historical production ranges for these factors
How this compares to traditional methods
This type of analysis is typically addressed through linear Root Cause Analysis (RCA) tools. However, traditional methods generally require the engineer to preselect a small sample of factors to investigate. This can easily introduce potentially misleading biases in their analysis which will impact the accuracy of results.
Our software uses a non-linear approach, much like your process operations. We’re able to evaluate all relevant factors, which improves insight and prediction accuracy. ExplainIt then delivers text-based explanations of the unusual factors that were observed, providing context to their unusualness and typical production ranges.
This entire process is performed in seconds with factor values automatically saved for your team to review, discuss or delve deeper to understand at a later time.
Shameless kudos here for all the people involved in this release
It takes a village to design, build, and test a product so this section will not do justice to all of the bright minds at Fero Labs who provided valuable input to this advanced market-first feature.
Huge thank you to everyone from Science and Engineering. With special mentions to Marcus Daly, Senior Research Engineer, and Will Brinkerhoff, Staff Engineer, for delivering value to our customers quickly and with care.
We’re incredibly proud of ExplainIt for Live Predictions, and would love an opportunity to show it to you live. Click here to book a demo with our team.
