When an operator has to make a decision within minutes, there's no time to analyze all the data in the factory. That's where Fero comes in. We built a "live optimization" feature that lets operators and engineers use Fero's machine learning-driven predictions and recommendations to guide real-time decision-making.
With live optimization enabled, the software's explainable machine learning models stream data directly from the production line, rather than relying on exported data sets. That means you get recommendations and predictions immediately and can apply them right away to production—maximizing cost savings. Within a month of enabling live optimization, one global steel company we work with saw saved as much as $4 per ton across several plants.
Given that this feature is fairly novel in both the artificial intelligence and manufacturing communities, we frequently get questions about how it works. Here are some of the most common questions:
How "live" is "live"?
Predictions arrive in seconds and optimization recommendations in under a minute. For virtually all use cases, this provides ample time for operators to make decisions.
You mentioned a steel use case. Will this work in my industry?
Our live optimization feature has demonstrated tangible results across the industrial world, from the chemical and CPG world to steel and cement manufacturing. All static models can easily be switched to live mode, helping you go from analyzing a static process to optimizing it in real time.
Does "live optimization" mean Fero directly controls my plant?
No. You can display recommendations to relevant engineers and operators and let them ultimately make the decisions. There also is an option to send recommendations directly to control systems, if you choose to do so. Ultimately, Fero software puts the knowledge in your hands—it's up to you how you apply it to optimize your production process.
Is "live" always better?
In most use cases we've seen, live optimization has dramatic results, increasing cost savings and yield and improving quality and processes. Of course, there are some cases where you might not need live predictions—you might simply want to understand the parameters affecting a given process, which don't change day-to-day. In that case, Fero's static mode is sufficient.
Additionally, it's best to wait to deploy a model live until your predictive accuracy is satisfactory. That's why we usually recommend testing Fero in static mode first, gaining confidence within the dataset, and then moving on to live optimization.
Besides cost savings, what are some other benefits?
Live optimization can help you improve on any benchmark that's important to your process, from increasing yield to achieving process stability. By working with real production values in real time, you'll be able to see how your current process can be improved and ensure you hit your KPIs.