Elevator Trapped-Event Detection Using Neural Network Features

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing elevator systems face challenges in accurately identifying real trapped events and optimizing rescue resource allocation, as many distress events are not caused by system failures and can be resolved without human intervention.

Innovation Solution

A method using a neural network model to analyze operational state data and generate a feature vector, determining the probability of a transport object being trapped in an elevator car, incorporating safety event data to enhance accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional alarm-based monitoring is used, then all distress events trigger rescue dispatches, but this leads to waste of rescue resources on non-trapped events

Engineering Contradiction:
Improveaccuracy of trapped event identificationVSAvoidrescue resource consumption
Core Design Contradiction:
ReliabilityVSLoss of energy

Solution Approach 1:

The patent transforms the monitoring approach from binary alarm detection to multi-dimensional parameter analysis. By collecting and analyzing multiple operational parameters (door status, car position, movement state, alarm status) simultaneously, the system changes the detection parameters from simple presence/absence to complex state combinations, enabling accurate differentiation between trapped events and false alarms

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces the traditional mechanical rescue dispatch system with an intelligent analysis system. Instead of automatically dispatching rescue personnel for all alarm events, the system uses data processing and pattern recognition to substitute human judgment, identifying which events truly require rescue intervention and which can be resolved automatically

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If comprehensive operational data is collected for analysis, then accuracy of trapped event detection is improved, but system complexity increases

Engineering Contradiction:
Improvedetection accuracy of trapped eventsVSAvoiddata collection and processing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent makes existing elevator system components serve multiple functions. The operational data collection system, originally designed for basic monitoring, is enhanced to simultaneously provide training data for machine learning models and real-time analysis for trapped event detection, eliminating the need for separate dedicated hardware systems

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system uses the elevator's own operational data and existing sensors to perform self-diagnosis and trapped event detection. By leveraging data already generated during normal operation (door status, position, movement), the system avoids requiring additional external monitoring equipment, making the complex analysis function self-contained within the existing infrastructure

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4585549A1Method and device for determining safety risk of elevator system
Publication Date: 2025.07.16 OTIS ELEVATOR CO
  • EP4585549A1 patent drawingFigure 1
  • EP4585549A1 patent drawingFigure 2
  • EP4585549A1 patent drawingFigure 3

AI summary

The present application relates to elevator technology and, in particular, to a method and device for determining a safety risk of an elevator system, and a non-transitory computer-readable storage medium storing a computer program for implementing the method. In accordance with an aspect of the present application, there is provided a method for determining a safety risk of an elevator system. According to the method, a device for determining the safety risk generates a feature vector based at least on operational state data of the elevator system, and subsequently determines a probability of a transport object being trapped in a car of the elevator system using a neural network model. In the above-described method, the feature vector and the probability are an input variable and an output variable of the neural network model, respectively, and the feature vector comprises components corresponding to combinations selected from a plurality of state values of a first category of features. Further, the state values of the first category of features are determined based on the operational state data.