Dynamic Leaf Guide Predictive Maintenance With Deep Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional predictive maintenance approaches for radiotherapy systems, particularly for dynamic leaf guides (DLGs) in linac systems, are complex, resource-intensive, and less adaptable due to the need for substantial domain knowledge and expertise, often leading to compromised performance and increased development costs.

Innovation Solution

A deep learning-based predictive maintenance system that trains on machine data to detect and diagnose DLG faults by classifying fault types and severities using a convolutional neural network (CNN), recurrent neural network (RNN), or long short-term memory (LSTM) network, enabling efficient and adaptable fault detection and diagnosis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional predictive maintenance approaches are used for DLG fault detection, then fault detection capability is achieved, but system complexity and development costs increase due to requiring substantial domain knowledge and expertise

Engineering Contradiction:

Improvefault detection capabilityVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces complex mechanical/domain-knowledge-based fault detection systems with an electronic deep learning-based electronic system. The deep learning model automatically learns fault patterns from operational data without requiring manual encoding of domain knowledge, thereby reducing system complexity while maintaining or improving fault detection capability.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The deep learning model performs self-learning from operational data to identify fault patterns. Instead of requiring external experts to manually program fault detection rules, the system autonomously develops its own diagnostic capabilities through training on historical data, reducing the need for substantial domain knowledge in system deployment.

Inventive Principle:

Principle #25Self-service

2Reliability

If conventional predictive maintenance approaches are used for DLG fault detection, then fault detection capability is achieved, but development time and costs increase

Engineering Contradiction:

Improvefault detection capabilityVSAvoiddevelopment time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The deep learning model is trained in advance on historical operational data from multiple DLGs to learn fault patterns before deployment. This preliminary training phase allows the model to be ready for immediate use without requiring time-consuming manual configuration or expert intervention during actual deployment, significantly reducing development time.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent uses operational data from multiple existing DLGs to train the deep learning model, effectively copying fault patterns from one system to another. This approach eliminates the need to manually program fault detection rules for each individual DLG, reducing development time and enabling rapid deployment across different machines.

Inventive Principle:

Principle #26Copying

3Reliability

If conventional predictive maintenance approaches are used, then fault detection is possible, but adaptability to different radiotherapy machines decreases

Engineering Contradiction:

Improvefault detection capabilityVSAvoidadaptability to different machines

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The deep learning model is designed to be universal across different DLG systems. By training on aggregated operational data from multiple DLGs with different configurations and operational characteristics, the model learns general fault patterns that apply across various machine types, enabling it to adapt to different radiotherapy machines without requiring machine-specific customization.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Reliability

If conventional predictive maintenance approaches are used, then fault detection is achieved, but maintenance efficiency decreases due to unnecessary servicing and costly breakdowns

Engineering Contradiction:

Improvefault detection capabilityVSAvoidmaintenance efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The deep learning model continuously monitors DLG operational data and provides real-time feedback on fault likelihood. This enables predictive maintenance by identifying faults before they occur, allowing maintenance to be performed only when necessary based on actual condition rather than following fixed schedules, thereby improving maintenance efficiency by reducing unnecessary servicing.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12377290B2Predictive maintenance of dynamic leaf guide based on deep learning

Publication Date: 2025.08.05 ELEKTA SHANGHAI TECH CO LTD

AI summary

Systems and methods for detecting and diagnosing faults in a radiotherapy system, such as a fault related to a dynamic leaf guide (DLG), are discussed. An exemplary predictive maintenance system includes a processor configured to receive machine data indicative of configuration and operation of a DLG in a target radiotherapy machine, apply a trained deep learning model to the received machine data, and detect and diagnose a DLG fault. The predictive maintenance system can train the deep learning model using data sequences constructed from the received machine data of the one more normal DLGs and the one or more faulty DLGs. Diagnosis of the DLG fault in the target radiotherapy machine includes classifying the DLG faults into different fault types or different fault severities.