Deep Learning Part Recognition for Elevator Maintenance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Field mechanics face challenges in efficiently identifying and determining damage in elevator parts due to numerous variations, challenging light conditions, and environmental factors, leading to increased time for maintenance and repair tasks.
Innovation Solution
A method using a camera on a mobile computing device for part recognition, employing supervised learning and deep learning models to classify elevator parts by capturing images, comparing them to CAD models and previous images, and determining damage through reconstruction error analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If field mechanics manually identify and assess elevator parts, then they can perform maintenance tasks, but the process is time-consuming and inefficient
Solution Approach 1:
The patent replaces manual mechanical inspection with an automated image recognition system using deep learning models. The system captures images of elevator parts with a camera and uses neural networks to automatically identify part types and detect damage, eliminating the need for manual visual inspection and significantly reducing time on site while maintaining high accuracy in part identification and damage assessment
Solution Approach 2:
The system enables self-service by allowing the mobile computing device to autonomously perform part identification and damage assessment without requiring expert mechanical knowledge. The deep learning model automatically processes images, compares them against trained data, and provides damage evaluations, making the maintenance process independent and highly efficient
2Measurement precision
If field mechanics manually inspect parts for damage, then they can identify issues, but challenging light conditions and environmental factors reduce accuracy
Solution Approach 1:
The patent replaces human visual inspection with machine vision technology that is immune to light conditions and environmental factors. The image recognition system processes digital images using deep learning algorithms that can accurately identify parts and detect damage regardless of lighting variations, dust, or other environmental challenges that hinder manual inspection
Solution Approach 2:
The system creates digital copies of physical parts through image capture and stores them in a database for comparison. These digital representations can be analyzed without being affected by environmental conditions, allowing consistent and accurate damage assessment by comparing captured images against stored reference images and CAD models
3Measurement precision
If the system uses deep learning models for part classification, then identification accuracy improves, but computational complexity increases
Solution Approach 1:
The patent segments the computational task by separating the deep learning model training and deployment into distinct phases. The system pre-trains classification and detection models using extensive datasets, then deploys these trained models on mobile devices for inference. This segmentation allows complex computational work to be done offline during model development, while field operations use the pre-trained models for fast, accurate part identification without requiring real-time training computational resources
Data Source
AI summary
According to one embodiment, a method of identifying a part of a conveyance system is provided. The method comprising: capturing an image of a part of a conveyance system using a camera; classifying the part of the conveyance system using supervised learning; and displaying a classification of the part of the part on a mobile computing device.