Earticle

초록

영어

This study proposes an artificial intelligence (AI)-based multi-output artificial neural network (ANN) classification model capable of automatically predicting optimal packaging methods and materials based on product characteristics. A total of 51 commercial meal-kit products distributed under ambient, chilled, and frozen conditions were analyzed to construct a comprehensive dataset comprising 19 packaging-related parameters such as distribution type, cooking requirement, packaging material, and sealing type. Of these, 17 parameters were used as input features, and two—packaging method and packaging material—served as output targets. The initial ANN model included an input layer of 17 nodes, two hidden layers with 128 neurons each (tanh activation), and two softmax output layers (nine and eight nodes for method and material classification, respectively). To optimize performance, Bayesian optimization was applied to five key hyperparameters: neuron count, learning rate, activation function, neuron division ratio, and optimizer type. The optimized model with four hidden layers and 1,024 neurons achieved classification accuracies of 85.8% for packaging method and 93.9% for packaging material, with corresponding F1-scores of 84.4% and 90.9%, respectively. Compared to the baseline, this represents a 22-44% improvement across metrics. The findings demonstrate the potential of AI-driven systems for intelligent, standardized, and sustainable packaging design, contributing to the advancement of smart manufacturing and eco-efficient food packaging automation

목차

Abstract
서론
재료 및 방법
1. 실험 재료
2. 밀키트 제품 분석을 위한 파라미터 정의
결과 및 고찰
1. 밀키트 제품 포장 방법 및 재질 분류 결과
2. 다중 출력 인공신경망 구조 최적화 결과
요약
감사의 글
참고문헌

키워드

저자

• 김태형 [ Tae Hyong Kim | 한국식품연구원 스마트제조연구단 ]
• 권기현 [ Kihyun Kwon | 한국식품연구원 스마트제조연구단 ]
• 김아나 [ Ah-Na Kim | 한국식품연구원 스마트제조연구단 ] Corresponding Author

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  한국포장학회지 [Journal of Korea Society of Packaging Science & Technology]

• 간기

  연3회

• pISSN

  1226-0207

• 수록기간

  1994~2026

• 등재여부

  KCI 등재

• 십진분류

  KDC 530 DDC 620