Dynamic Evacuation Strategy Using Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing evacuation strategies in production facilities are statically specified and lack dynamic adaptation to real-time situations, such as fires or hazardous substance leaks, which can lead to inefficient evacuation processes.
Innovation Solution
A method utilizing an artificial neural network trained with topology and event data to simulate the temporal behavior of people and events within a building, allowing for the determination of a dynamic evacuation strategy based on sensor inputs, including sensor positions, weather, occupancy, and machine status, to efficiently direct evacuees.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static evacuation scenarios are used, then evacuation can be initiated directly, but the evacuation strategy cannot adapt to real-time situations
Solution Approach 1:
The patent transforms static evacuation scenarios into dynamic ones by implementing real-time sensor monitoring that detects current conditions (smoke, heat, occupancy) and continuously updates evacuation strategies. The system transitions from predetermined fixed paths to adaptive routes that respond to changing environmental conditions, resolving the contradiction between reliability through adaptability and system complexity.
Solution Approach 2:
The patent implements feedback loops where sensor data from the building environment is continuously fed back to the evacuation control system. This feedback mechanism allows the system to monitor real-time conditions and adjust evacuation strategies dynamically, enabling adaptability while managing complexity through automated closed-loop control.
2Reliability
If expert-developed evacuation strategies are used, then strategies can be verified through simulations, but the strategies lack real-time adjustment capability
Solution Approach 1:
The patent applies preliminary action by pre-training artificial intelligence models with simulated evacuation scenarios and expert strategies during peacetime. This pre-computed knowledge base enables rapid real-time decision-making during actual emergencies, resolving the contradiction between strategy effectiveness from expert verification and response time by having solutions prepared in advance.
Solution Approach 2:
The patent uses copying by creating virtual replicas of the building and occupancy scenarios through simulation. These copied models allow expert strategies to be tested and refined without affecting real operations, and the validated strategies are then applied to actual evacuation situations, maintaining effectiveness while enabling rapid response.
3Productivity
If traditional evacuation methods are used, then implementation is straightforward, but efficiency is reduced due to lack of dynamic adaptation
Solution Approach 1:
The patent replaces traditional mechanical evacuation control systems with an artificial neural network-based intelligent system. This substitution enables dynamic adaptation and optimization of evacuation routes based on real-time conditions, significantly improving evacuation efficiency despite the increased computational complexity, which is managed through automated training processes.
Data Source
Figure 1~4
Figure 5~7
AI summary
The invention relates to a method for determining an evacuation strategy for the evacuation of a building, comprising the following steps: receiving (101) sensor signals representing an event requiring the evacuation of the building, detected by a sensor encompassed by the building; determining (103) an evacuation strategy for evacuating from the building using an artificial neural network based on the detected event; and outputting (105) evacuation strategy signals representing the determined evacuation strategy. The invention further relates to a method for training an artificial neural network, a device, an artificial neural network, a computer program, and a machine-readable storage medium.