Abstract: A universal deep probabilistic decision-making framework for dynamic process modeling and control, referred to as Deep Probabilistic Decision Machines (DPDM), is presented. A predictive model enables the generative simulations of likely future observation sequences for future or counterfactual conditions and action sequences given the process state. Then, the action policy controller, also referred to as decision-making controller, is optimized based on predictive simulations. The optimal action policy controller is designed to maximize the relevant key performance indicators (KPIs) relying on the predicted experiences of sensor and target observations for different actions over the near future.

Abstract: An AI-based anomaly signatures warning recommendation system is provided. The system includes a memory having computer-readable instructions stored therein and a processor configured to execute the computer-readable instructions to access a multi-asset connected system having a plurality of production and/or process lines. Each of the plurality of production lines includes a plurality of assets. The processor is configured to access production data corresponding to a plurality of products manufactured in each of the plurality of production lines and to access sensor signal data corresponding to each of the plurality of assets. The sensor signal data is indicative of health of each of the plurality of assets. The processor is further configured to process the production data and sensor signal data for each of the plurality of assets to identify one or more anomaly instances and to perform similarity analysis on the one or more anomaly instances to identify one or more anomaly signatures.

Abstract: An artificial intelligence (AI)-based defect diagnosis system for automatic identification of one or more defect drivers in a manufacturing environment is presented. The diagnosis system includes an input data module, an input specifications module, a product selection module, a product grouping module, a defect driver identification module, and an output module. A related method is also presented.

Abstract: Methods and systems for controlling production resources in a supply chain are described. The system automatically generates predicted supply chain operational metrics across a nodes of a supply chain. The system automatically infers causal factors that impact the predicted supply chain operational metrics. The causal factors include a change to a utilization of the production resource. The system communicates a user interface including production runs being scheduled on the production resource including a user interface element representing the scheduling of the production run associated with a value at risk. The system receives input causing a change to the utilization of the production resource. The change to the utilization of the production resource impacts the predicted supply chain operational metrics including the value at risk associated with the scheduling of the production run.

Abstract: An AI-based smart asset health surveillance system for a connected system is presented. The connected system includes a plurality of production and/or process lines, wherein each of the plurality of production and/or process lines includes a plurality of assets. The smart asset health surveillance system includes a memory having computer-readable instructions stored therein; and a processor configured to execute the computer-readable instructions. The processor is configured to execute the computer-readable instructions to monitor the plurality of assets and to automatically predict one or more downtime and/or anomalous events for the plurality of assets. An AI-based smart asset health surveillance method is also presented.

Abstract: A spurious outlier detection-system is provided. The system includes a memory having computer-readable instructions stored therein and a processor configured to execute the computer-readable instructions to receive time-series data from one or more sensors and/or applications, process the time-series data to detect one or more change points based on a pre-defined cost function. The processor is configured to identify data chunks between the change points using pre-determined window sizes and to estimate smooth reconstructed values (SRVs) for each of the change point data chunks between two consecutive change points to identify one or more global outliers from the SRVs. The processor is configured to determine distribution of the global outliers using kernel density for each change point data chunk and identify one or more true outliers from the distribution of the global outliers based upon a skewness of the distribution.

Abstract: A system is presented for predicting the amount of energy required by a manufacturing plant over a period of time, and then providing a recommendation on the best amount of energy to sell at each period, taking into consideration the energy requirements based on the production schedule, the price of energy, possible penalties for over selling energy, and other factors. A user interface is provided to present relevant information to a user, allow the user to plan the use of energy, and determine the energy offering for sale in each period. Based on a known future production schedule of a steel mill, implementations provide a forecast of electrical energy usage per hour over the same time horizon as the production schedule. Based on the forecast and electricity selling and billing rules, a recommended amount of energy to be sold to the market by the hour is presented with the corresponding offer price.

Abstract: Methods and systems to predict a supply chain performance are described. A system receives supply chain data for delivery of a product. The supply chain data includes input signals comprising operational plans and observed supply chain operational metrics. The input signals include a delivery date of the product. The system automatically generating predicted supply chain operational metrics across including a value at risk that is predicted for the product. The system automatically infers causal factors that impact the predicted supply chain operational metrics including impacting the value at risk that is predicted for the product. The system automatically generates action recommendations for the supply chain. An action recommendation includes a first predicted value impact and a sequence of actions impacting the product the delivery date of the product and the value at risk that is predicted for the product.

Abstract: Methods and systems for controlling inventory in a supply chain are described. The system receives supply chain data including input signals comprising operational plans and observed supply chain operational metrics. The system automatically generates predicted supply chain operational metrics including a value at risk that is predicted for a product. The system automatically infers causal factors including a shipment of the product. The causal factors impact the predicted supply chain operational metrics. The system communicates a user interface for shipments of the product and the system receives input causing a change to a shipment of the product impacting the predicted supply chain operational metrics including the value at risk for the first product.