Scholarlink

<p>The light emitting diode (LED) bulbs, lamps, and tubes are essential in the lighting market, and a comprehensive life cycle assessment remains pivotal for testing these products. Industry standards mandate that samples undergo continuous operation for a minimum of 6000&#xa0;h, which can be costly for manufacturers if they change manufacturing processes. To address this issue, a dataset, namely the Lumen_Maintenance_Temperature_Humidity dataset, is created by continuously turning on samples from four different manufacturers for approximately 12,000&#xa0;h in tropical weather conditions. Machine learning (ML) models, multivariate polynomial regression, k-nearest neighbor regression, and random forest regression predict how well pcLEDs maintain brightness based on data from lifespan tests, including how long they have been running, the surrounding temperature, and humidity. Additionally, this attempt uses several hyperparameters to estimate lumen maintenance using the newly created dataset. The success of ML models in predicting lumen maintenance from sixteen smaller datasets has been measured using mean squared error (MSE) and <i>r</i><sup>2</sup> scores. It has been observed that the random forest regressor performed better than the other models in predicting the lumen maintenance of all the samples and achieved up to a maximum of 99.98 test <i>r</i><sup>2</sup> score. However, for two samples from manufacturers 2 and 4, the multivariate polynomial and k-nearest neighbor regressor models excel over the random forest regressor and achieves up to a maximum of 99.78 in the test <i>r</i><sup>2</sup> score. This research is applicable not only to India but also to tropical countries worldwide, allowing manufacturers to apply findings and modify industry standards accordingly.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting lumen maintenance of phosphor-converted light emitting diodes in Indian tropical conditions using machine learning

  • 期刊论文

  • 发表日期：

    2025年5月19日

查看全文
  • Arindam Chakraborty,
  • Chiradeep Mukherjee,
  • Rajiv Ganguly,
  • Monojit Mitra

摘要

The light emitting diode (LED) bulbs, lamps, and tubes are essential in the lighting market, and a comprehensive life cycle assessment remains pivotal for testing these products. Industry standards mandate that samples undergo continuous operation for a minimum of 6000 h, which can be costly for manufacturers if they change manufacturing processes. To address this issue, a dataset, namely the Lumen_Maintenance_Temperature_Humidity dataset, is created by continuously turning on samples from four different manufacturers for approximately 12,000 h in tropical weather conditions. Machine learning (ML) models, multivariate polynomial regression, k-nearest neighbor regression, and random forest regression predict how well pcLEDs maintain brightness based on data from lifespan tests, including how long they have been running, the surrounding temperature, and humidity. Additionally, this attempt uses several hyperparameters to estimate lumen maintenance using the newly created dataset. The success of ML models in predicting lumen maintenance from sixteen smaller datasets has been measured using mean squared error (MSE) and r2 scores. It has been observed that the random forest regressor performed better than the other models in predicting the lumen maintenance of all the samples and achieved up to a maximum of 99.98 test r2 score. However, for two samples from manufacturers 2 and 4, the multivariate polynomial and k-nearest neighbor regressor models excel over the random forest regressor and achieves up to a maximum of 99.78 in the test r2 score. This research is applicable not only to India but also to tropical countries worldwide, allowing manufacturers to apply findings and modify industry standards accordingly.