Using Artificial Intelligence in Smart Products
As introduced in previous editions of this newsletter, Artificial Intelligence is a powerful new tool in leveraging technology, allowing us to rethink and optimize the way we use to process and analyze data. Artificial Intelligence helps us to offer smarter systems in almost all business areas. At Arvato Systems, we use the latest artificial intelligence technology in Advanced Analytics to create smart products, streamline processes as well as to automate decision making systems and services. Thus, we focus on improvement potential as well as new opportunities which are opening up by leveraging the full potential of AI. Our latest proof of concept (POC) is based on AI and focuses on variations of prediction models as well as classification systems i.e., generic text classification and video or image recognition.
In this recent POC we used AI to implement an autonomous prediction service with built-in anomaly detection. How does the detection service work? Customers send their time-series data continuously with reference to a unique key which provides our service the required customer information. Our service then uses the received data to automatically detect the content and categorizes it. Additionally, it creates a prediction model in real-time based on the historical data of the customer. AI is also used for the forecasting process as the POC automatically detects several parameters which are used for a detailed forecasting, including seasonal or general trends. As a result, our customers can use the prediction service to, e.g. optimize pricing strategies. The customer can also use the prediction service to act and react to price changes within the market very quickly, and thus, increase positive business impacts. Another use case for the prediction service is to support predictive maintenance. The forecast results when combined with the anomaly detection can help to catch anomalies in real-time. Such abnormalities could occur in specific sensors within large industrial systems. Automatic detection of these anomalies can return an alert to a service employee who is responsible for maintenance, allowing the defect to be resolved before it becomes a business critical issue.
Within the trained system, the customer can add new data points while the backend compares the received data with its previously calculated forecasted data in real-time. The comparison enables the system to detect anomalies and outliers within the data set. The outcomes of the data analysis are then returned to the customer directly, which can be used, e.g. for system optimizations and predictive maintenance.
In a further step, our smart service POC was also integrated with the Arvato Systems Smart Energy Platform (SEP), which was introduced in more detail in our may newsletter (read more here). We provide a full cloud-based integration of AI and anomaly detection service into the SEP, helping to “smartify” the processing and analyses of any kind of IoT sensor data. In our cloud, we offer several micro-services located in our application layer. These include basic insights and analyses as well as fully automated and intelligent decision systems. In a separate layer, we can store big data, which can be pre-processed with our transformation engine and then transferred to the micro-services and visualization processes.
For more information please contact Ricardo R. Gameiro.
