Scholarlink

<p>This work discusses the fabrication of microelectrode arrays (MEAs) using a type of machining process called Reverse-Micro-Electrical-Discharge Machining (Reverse-µEDM). The main aim here is to forecast and fine-tune the surface roughness because MEAs hold a crucial role in the health field for gathering neural signal fields i.e., highly accurate and in the optimal range. MEAs were made with Reverse-µEDM, known for its better accuracy and precision. Using Taguchi’s L18 test plan, the effect of main factors (voltage, capacitance, and feed rate) was checked carefully. ANOVA showed where variation came from: most of it was from capacitance (86%), then voltage (11%), and feed rate gave a tiny 0.2% contribution to surface roughness. Surface roughness was measured with non-contact profilometer; it took parameters for roughness on surfaces. A Scanning Electron Microscopy (SEM) was used analyze the topography and microstructure of MEAs. Surface roughness has been predicted using machine learning models such as Artificial Neural Network (ANN), Support Vector Regression (SVR), Random Forest Regression, and Gradient Boosted Trees along with Gaussian Process Regression (GPR). Among them, Random Forest gave the highest accuracy for prediction of surface roughness with maximum value of R² and low error measured by Mean Absolute Error (MAE = 0.18&#xa0;μm). The Neural Network was almost equally good but took a longer time to train. SVR and GPR provided medium accuracy; Gradient Boosting performed poorly in this small dataset. Leave-One-Out Cross-Validation (LOOCV) yielded strong performance estimates given the small amount of data.</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

Predictive modeling and optimization of surface roughness in Reverse-µEDM fabricated microeletrode arrays using ML models

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月6日

查看全文
  • Suresh Pratap,
  • Prakash Kumar,
  • Hreetabh Kishore,
  • Somak Datta,
  • Abdin Bedada Huluka

摘要

This work discusses the fabrication of microelectrode arrays (MEAs) using a type of machining process called Reverse-Micro-Electrical-Discharge Machining (Reverse-µEDM). The main aim here is to forecast and fine-tune the surface roughness because MEAs hold a crucial role in the health field for gathering neural signal fields i.e., highly accurate and in the optimal range. MEAs were made with Reverse-µEDM, known for its better accuracy and precision. Using Taguchi’s L18 test plan, the effect of main factors (voltage, capacitance, and feed rate) was checked carefully. ANOVA showed where variation came from: most of it was from capacitance (86%), then voltage (11%), and feed rate gave a tiny 0.2% contribution to surface roughness. Surface roughness was measured with non-contact profilometer; it took parameters for roughness on surfaces. A Scanning Electron Microscopy (SEM) was used analyze the topography and microstructure of MEAs. Surface roughness has been predicted using machine learning models such as Artificial Neural Network (ANN), Support Vector Regression (SVR), Random Forest Regression, and Gradient Boosted Trees along with Gaussian Process Regression (GPR). Among them, Random Forest gave the highest accuracy for prediction of surface roughness with maximum value of R² and low error measured by Mean Absolute Error (MAE = 0.18 μm). The Neural Network was almost equally good but took a longer time to train. SVR and GPR provided medium accuracy; Gradient Boosting performed poorly in this small dataset. Leave-One-Out Cross-Validation (LOOCV) yielded strong performance estimates given the small amount of data.