Prescriptive analytics is the process of analyzing historical operations data, creating machine learning models with that information and determining the best course of action for a specific scenario. Data collected during asset performance monitoring contains enormous amounts of information about how assets operate. Machine learning tools can mine historical operating data for patterns in asset performance. These artificial intelligence models can then extrapolate from past performance to simulate what could happen under different scenarios.
Prescriptive analytics allows plant operators to input desired outcomes and receive recommendations about how to change the manufacturing processes to achieve those goals. The cost of shutting down production is estimated to be nearly $3 million/ hour, so any method, like prescriptive analytics, that lets plant operators experiment without risking production gives companies a huge advantage.
Other potential benefits of using prescriptive analytics tools:
- Can provide you entirely new ways to optimize assets
- Service life can be extended by adjusting the operating conditions of an asset
- Different combinations of equipment settings can be explored without affecting production
How prescriptive analytics works
Prescriptive analytics uses statistical and machine learning models to provide organizations with information about how to achieve optimal asset performance. Enormous amounts of data are fed to a specialized computer algorithm that applies machine learning to asset performance monitoring records. The artificial intelligence algorithm finds any connections between various operating conditions by looking for patterns and correlations in the data.
Going one step further than predictive analytics, prescriptive analytics differs from other methods of simulating industrial processes by not requiring a deep practical understanding of the machinery and equipment being replicated. Previous methods for creating “digital twin” simulations demanded the most skilled operations and design personnel to spend time away from the floor working with programmers, an enormous drain on human resources. With machine learning methods, a general-purpose algorithm can be trained on a wide range of processes with very little human input.
What prescriptive analytics can do
Prescriptive analytics gives reliability management teams the opportunity to digitally experiment with process adjustments without actually risking manufacturing downtime. The software models created with prescriptive analytics algorithms offer a way to explore what-if scenarios to increase efficiency and asset optimization in the safety of a software simulation.
With complex industrial processes, the volume of choices at every manufacturing step can be overwhelming even to skilled technicians, and reliability management often defaults to isolated incremental changes. This can prevent the discovery and implementation of more efficient equipment configurations that require multiple coordinated tweaks. With tightening profit margins, few companies are willing to approve more than a few small changes to critical infrastructure at the same time. Unless all the changes required to shift to a more efficient method of operation are all individually more efficient, there will likely be significant financial pushback. Prescriptive analytics models allow these coordinated shifts to be tested and explored safely in digital environments by asset performance management teams.
This process can be reversed as well; prescriptive analytics models can be provided with desired outcomes and return recommendations about how to achieve those goals. Instead of adjusting the parameters of the simulation and seeing how it changes the output, the model can be instructed to change its parameters based on desired performance. For instance, a prescriptive analytics model can be told to reduce energy usage while keeping production constant, or to find a way to safely delay maintenance until a certain date. The software will then backtrack from the desired behavior and provide recommendations about how it can be achieved.
Making the most of prescriptive analytics
There may be serious resistance from personnel to adopt prescriptive analytics methods. After all, reliability management is primarily concerned with minimizing risk. The results returned by prescriptive analytics models may be counterintuitive to a technician that has extensive theoretical knowledge about how the equipment should work. Ensuring that everyone at an enterprise is committed to using prescriptive analytics is an important first step.
Ideally, prescriptive analytics should be integrated with the asset performance management software being used at an enterprise. For many organizations, having to turn away from the current conditions of the plot to boot up another piece of software to explore what-if scenarios can be a huge impediment to truly integrating prescriptive analytics.
What is the difference between predictive and prescriptive analytics?
Predictive analytics uses machine learning and artificial intelligence to analyze historical sensor and service data to find failure signatures that humans may miss and predict when failures will reoccur. Prescriptive analytics takes this a step further and finds patterns in normal performance that may be of interest to plant operators and provides recommendations on changes to make to alter performance outcomes.
What is prescriptive modeling?
Prescriptive modeling is when prescriptive analytics are used to simulate different outcomes. A simulation of an industrial process is built using machine learning, and the resulting model can be used to test different operating conditions.