Analisis Perbandingan Kinerja Sensor VL53L0X dan HC-SR04 pada Sistem Pencatatan Stok Barang dengan Support Vector Regression

Abstract

Pencatatan stok barang secara manual di departemen Production Planning and Inventory Control (PPIC) pabrik farmasi sering menghasilkan data yang tidak akurat dan menghambat penentuan Reorder Point (RoP). Penelitian ini bertujuan merancang sistem pencatatan stok otomatis berbasis Internet of Things (IoT) menggunakan mikrokontroler ESP32-S3 untuk memantau ketersediaan barang secara real-time. Penelitian ini melibatkan analisis perbandingan kinerja sensor jarak VL53L0X berbasis Time of Flight (ToF) dan sensor ultrasonik HC-SR04. Karena sensor HC-SR04 sangat rentan terhadap perubahan suhu lingkungan, algoritma Support Vector Regression (SVR) diterapkan dengan memanfaatkan data suhu dari sensor BMP280 untuk memodelkan dan mengoreksi pembacaan jarak yang bersifat non-linear. Berdasarkan hasil evaluasi, pembacaan sensor HC-SR04 mentah menghasilkan tingkat kesalahan tertinggi dengan nilai Root Mean Square Error (RMSE) 2,54 cm dan Mean Absolute Error (MAE) 2,42 cm. Sementara itu, sensor VL53L0X terbukti jauh lebih stabil dengan nilai RMSE 0,30 cm. Namun, penerapan algoritma SVR pada sensor HC-SR04 berhasil menekan tingkat kesalahan secara signifikan hingga mencapai akurasi paling optimal dengan angka RMSE 0,10 cm dan MAE 0,10 cm. Hasil penelitian ini membuktikan bahwa algoritma SVR secara efektif mampu mengompensasi kelemahan sensor ultrasonik akibat gangguan lingkungan, sehingga sistem mampu menghasilkan data kuantitas stok yang akurat dan memicu mekanisme notifikasi RoP visual secara otomatis pada website. Sebagai pengembangan selanjutnya, sistem ini disarankan untuk mengintegrasikan fitur kalibrasi dimensi barang secara otomatis serta algoritma analisis time-series untuk memprediksi estimasi waktu kekosongan stok berdasarkan tren historis.

Manual inventory recording in the Production Planning and Inventory Control (PPIC) department of a pharmaceutical factory often produces inaccurate data and hinders the determination of the Reorder Point (RoP). This study aims to design an Internet of Things (IoT)-based automatic inventory recording system using an ESP32-S3 microcontroller to monitor inventory availability in real-time. This study involves a comparative analysis of the performance of the Time of Flight (ToF)-based VL53L0X proximity sensor and the HC-SR04 ultrasonic sensor. Since the HC-SR04 sensor is highly susceptible to changes in environmental temperature, the Support Vector Regression (SVR) algorithm is applied by utilizing temperature data from the BMP280 sensor to model and correct non-linear distance readings. Based on the evaluation results, the raw HC-SR04 sensor readings produce the highest error rate with a Root Mean Square Error (RMSE) of 2.54 cm and a Mean Absolute Error (MAE) of 2.42 cm. Meanwhile, the VL53L0X sensor is proven to be much more stable with an RMSE of 0.30 cm. However, the application of the SVR algorithm to the HC-SR04 sensor succeeded in significantly reducing the error rate to achieve the most optimal accuracy with an RMSE of 0.10 cm and an MAE of 0.10 cm. The results of this study prove that the SVR algorithm is effectively able to compensate for the weaknesses of the ultrasonic sensor due to environmental disturbances, so that the system is able to produce accurate stock quantity data and trigger a visual RoP notification mechanism automatically on the website. For future work, this system is recommended to integrate an automatic item dimension calibration feature, as well as a time-series analysis algorithm to predict the estimated time of stock depletion based on historical trends.