Smart System Learns Industrial Machine Behavior

This patent describes a smart system that uses sensors and machine learning to predict outcomes like production volume or maintenance needs for industrial machines, based on industry-specific data.

What it covers

This patent covers a system designed for industrial settings. It uses multiple sensors to gather data about a machine's operation. A central controller then uses a machine learning component to analyze this data. This analysis helps the system learn patterns that predict specific outcomes, such as the machine's reaction rate, how much it produces, or if it needs maintenance. Crucially, the machine learning starts with a pre-built model, informed by industry feedback, and can also learn to select the most useful sensors for its predictions. For example, it could predict when a specific manufacturing machine will require a part replacement based on its vibration and temperature readings.

What it doesn't cover

• —Systems that do not use machine learning to analyze data patterns.
• —Systems that do not learn predictive patterns for industrial machine outcomes.
• —Systems that do not use industry-specific feedback to seed their learning models.
• —Systems that do not learn a preferred subset of input sensors.
• —Systems predicting outcomes unrelated to industrial machines, like weather.

The clever bit

The core innovation is seeding the machine learning model with industry-specific feedback and learning a preferred subset of sensors. This means the system doesn't start from scratch; it's guided by prior knowledge relevant to the specific industrial context, making its predictions more accurate and efficient from the outset.

Why it matters

This patent is relevant to the growing field of Industrial Internet of Things (IIoT). It provides a framework for making industrial processes more efficient and predictable by leveraging data analytics and machine learning. Such systems can help reduce downtime and optimize resource allocation in factories and other industrial environments.

Real-world examples

• 1.Predictive maintenance systems in factories
• 2.Process optimization in chemical plants
• 3.Production volume forecasting in manufacturing
• 4.Smart energy grid management