AI System for Diagnosing Machines and Managing Data in Factories

This patent describes a system that uses artificial intelligence to detect problems in industrial machines while smartly adjusting how much data it collects and sends to avoid overwhelming the network.

What it covers

The system collects data from various sensors connected to industrial machines, like a robot's vibration sensors or a factory line's temperature gauges. A 'data collector' gathers this information from a selected group of these sensors, called 'input channels,' based on a specific routine (Claim 1). An 'expert system analysis circuit' then uses a type of artificial intelligence called a 'neural network' (which could be probabilistic, time delay, or convolutional, as per Claim 1) to analyze these data streams and identify 'fault conditions' or problems with the machine components (Claim 1). Crucially, the system also monitors the total amount of data being collected, called the 'aggregate rate' (Claim 1). If this rate goes over the network's current capacity, known as the 'bandwidth allocation rate,' the system automatically asks the network for more bandwidth (Claim 1). Until more bandwidth is available, it can temporarily increase data storage capacity (Claim 1), or it might reduce the data by eliminating some collected information (Claim 2), deactivating monitoring points (Claim 4), or reducing how often data is sampled or its detail level (Claim 6). For example, it could detect an unusual vibration pattern in a motor and, if the network is busy, temporarily reduce the sampling rate of less critical sensors to ensure the critical vibration data gets through.

What it doesn't cover

• —Does not cover systems that diagnose machine faults without using a neural network as part of their expert system analysis circuit.
• —Does not cover general data collection systems that do not dynamically monitor and request changes to network bandwidth allocation rates.
• —Does not cover systems that only detect faults but do not implement specific data reduction strategies like eliminating data, deactivating monitoring points, or modifying sampling parameters when bandwidth is exceeded.
• —Does not cover systems operating outside of an industrial environment or not focused on machine components.
• —Does not cover systems that simply drop data without first attempting to increase bandwidth allocation or manage data capacity.

The clever bit

The truly clever part is combining AI-driven fault detection with an adaptive system that actively manages network bandwidth and data collection in real-time. It doesn't just detect issues; it also ensures the data needed for detection can actually get where it needs to go, even under network strain, by intelligently requesting more bandwidth or reducing less critical data.

Why it matters

This technology is important because it helps factories prevent expensive machine breakdowns by catching problems early using AI. By intelligently managing data flow, it ensures that critical diagnostic information can be sent and analyzed even when network resources are limited, which is common in complex industrial settings. This capability can significantly reduce downtime and maintenance costs, making industrial operations more efficient and reliable.

Real-world examples

• 1.Predictive maintenance systems in smart factories
• 2.Industrial IoT platforms for equipment monitoring
• 3.Condition monitoring systems for manufacturing robots
• 4.Sensor networks in power plants or refineries
• 5.Automated diagnostics for heavy machinery