Predictive Analytics: Predictive and Prescriptive Technology with AI and Machine Learning
At Wonderware Iberia, we offer Predictive Analytics (formerly called PRiSM), a market-leading software for implementing a predictive and prescriptive strategy.
Thanks to automation in the industry, there is an increasing number of equipment with multiple sensors in plants and infrastructures, which allow controlling production processes. These sensors are producing huge amounts of data, which are stored in Historians and transformed into contextualized, useful, and valuable information by combining the data with Artificial Intelligence tools to predict the behavior of the equipment.
Currently, one of the objectives of the maintenance departments is to be more proactive and anticipate future equipment failures through the use of predictive strategies. In this way, it will be possible to carry out maintenance actions only when the equipment requires it and not based on fixed periods of time or usage statistics.
The application of predictive tools allows organizations to avoid breakdowns and unscheduled shutdowns and save costs associated with preventive maintenance and planned shutdowns for such interventions, which are often unnecessary. However, the prediction of future failures does not provide all the potential by itself; it requires specific prescriptive follow-up actions to obtain the expected value.
At Wonderware Iberia, we offer Predictive Analytics (formerly called PRiSM), a market-leading software for the implementation of a predictive and prescriptive strategy.
Next, we will briefly explain the functionalities and main features of Predictive Analytics.
I how Does Predictive Analytics Achieve Early Failure Detection?
Predictive Analytics uses a patented algorithm called OPTiCS that is based on advanced pattern recognition (Advanced Pattern Recognition) and machine learning technology. In addition, it has algorithms such as ICA (Independent Component Analysis) and LSH (Locality-Sensitive Hashing), as well as the possibility of importing algorithms external to the software. For systems with lower levels of historical repeatability, high noise, or process systems, Predictive Analytics uses a plugin for a predictive algorithm called KANN. This algorithm allows the user to create models that predict future values for the signals. The algorithm uses artificial neural network technology and allows users to create operational profiles with a specific set of inputs and outputs and test how the outputs will evolve in the future through data playbacks.
Predictive Analytics learns the particular operating profile of an asset during all load, environment, and operating process conditions. The historical data from the equipment’s sensors is entered into the software’s modeling process, and advanced algorithms generate the equipment’s operational profile. Once the model is deployed, it is compared with the real-time operating data to detect and alert subtle deviations from the expected behavior of the equipment. Once a problem has been identified, the software can help in the root cause analysis and provide diagnostics to help the user understand the reason and importance of the problem.
I Advanced Pattern Recognition (APR)
Predictive Analytics is a software based on the modeling of equipment behavior through advanced pattern recognition (APR). This modeling is based on the historical data of the equipment itself that describes its behavior exactly, previously cleaning the anomalous behaviors from the history.
I Data Sources and Integration
Predictive Analytics integrates with a wide variety of data historian systems, control and monitoring systems, and can be implemented on-premise or in the cloud. The system is highly scalable and can be used to monitor a single asset, a plant, or hundreds of remote assets in multiple plants. The results of PRiSM models can be easily integrated with other business systems through the use of web services and the available RESTful API.
I Creation of Predictive Models
PRiSM Client is a desktop-based application that is used to develop, train, validate, and implement equipment models and their respective alert notifications. It is equipped with templates and a database of assets and known conditions that streamline the model creation process, making it easier for users to create and maintain their own models. The intuitive and graphically driven process allows models to be built in minutes rather than days or weeks and does not require any programming or detailed knowledge of the equipment.
I Monitoring and Data Analysis on the Web
The web-based application, PRiSM Web, is used to manage alerts, quickly retrain models, and analyze and graph model results. It organizes alarm information in a hierarchical structure that allows users to identify systems that are in an abnormal state and then view the individual components that generate the alarm for further analysis.
I Alerts and Notifications
Users can set alert thresholds to communicate when the deviation between actual values and predicted values exceeds the allowed limits. Alerts can be managed in several ways, including by category, level, criticality, duration, and frequency. Each alert event is also directly linked to a graph for that asset showing the event data, threshold limits, and times when the values are in alarm. Relevant users and groups can be notified in real time if an asset is in an alert state through the generation of customizable email notifications.
I Data Analysis and Fault Diagnosis
The software includes a variety of advanced model-based and statistical comparison applications, as well as Business Intelligence tools that allow users to spend less time searching for potential problems. Users have the ability to view raw modeling data, model results, compare the performance of similar assets of the same type, and view the effects of alerts. Statistical applications interpret the data through visual representations so that organizations do not require data scientists and expert engineers on the equipment to interpret the results.
Predictive Analytics is equipped with fault diagnostic capabilities to help the user determine the cause of the identified abnormality and how to avoid it in the future. The diagnostic capability eliminates the likelihood of an engineer attributing abnormal operating conditions to the wrong variable.
I Prescription of Maintenance Actions
Predictive Analytics is more than a predictive analysis tool thanks to the ability to relate prescriptive maintenance according to the deviation of the behavior. Once the software has detected and notified an alert due to the deviation in the behavior of the equipment, it performs a fault diagnosis and indicates what maintenance actions should be carried out to prevent the failure.
The software allows storing all the technical knowledge and the experience of the maintenance team in the fault diagnoses, so that once the maintenance actions to be carried out have been determined, they can be repeated in the future through the prescriptive maintenance of the solution. This allows organizations to standardize the methodology for resolving deviations in behavior and guide technical personnel towards the possible cause of the future failure.
These are just the main functionalities of the Predictive Analytics solution, however, there are many more that allow organizations to extract the maximum return on investment from their assets, ensuring that they do not suffer unscheduled shutdowns and minimizing preventive maintenance as much as possible.
If you want to know more about Predictive Analytics or we can help you resolve any questions, do not hesitate to contact us.
