Pradhan Sachin, Bhattarai Jaya Sharma, Murugavel Muthuchamy, Sharma Om Prakash
Department of Chemistry, School of Basic Sciences, Shri Ramasamy Memorial University Sikkim, fifth Mile, Tadong, Gangtok 737102, East Sikkim, India.
School of Information Technology, Shri Ramasamy Memorial University Sikkim, fifth Mile, Tadong, Gangtok 737102, East Sikkim, India.
ACS Omega. 2025 Apr 18;10(16):16597-16601. doi: 10.1021/acsomega.5c00027. eCollection 2025 Apr 29.
Many scientific and industrial applications depend on the precise measurement of chemical concentrations. The current study demonstrates how an inventive method of combining photographic images with a machine learning (ML) model successfully estimates the concentration of a chemical compound in solution. A machine learning model using linear regression with L2 regularization (ridge regression model) was developed as a part of a predictive model. The model was trained on captured images of KCrO solutions following the standard setup. After completing the training, the model was evaluated using a data set of test samples. The prediction precision of the model had been evaluated using 210 images and a high correlation between actual and predicted KCrO concentrations was obtained with MAE, MSE, and RMSE of 1.4 × 10, 3.4 × 10, and 1.0 × 10, respectively. The ridge regression model is also extended to predict the concentration of potassium permanganate (KMnO) and highlights the potential of integrating machine learning techniques with image analysis to accurately quantify the concentration of any chemical species in the solution state. As this model depends solely on the color intensity of the sample without any molecular interactions, it exceeds the limitations of the Beer-Lambert law. The created machine learning model also minimizes the requirement of substantial expertise and training and hence bridges the gap between experienced and novice analysts.
许多科学和工业应用都依赖于化学浓度的精确测量。当前的研究展示了一种将摄影图像与机器学习（ML）模型相结合的创新方法如何成功估计溶液中化合物的浓度。作为预测模型的一部分，开发了一种使用带有L2正则化的线性回归的机器学习模型（岭回归模型）。该模型按照标准设置在捕获的KCrO溶液图像上进行训练。训练完成后，使用测试样本数据集对模型进行评估。使用210张图像评估了模型的预测精度，实际和预测的KCrO浓度之间具有高度相关性，平均绝对误差（MAE）、均方误差（MSE）和均方根误差（RMSE）分别为1.4×10、3.4×10和1.0×10。岭回归模型还被扩展用于预测高锰酸钾（KMnO）的浓度，并突出了将机器学习技术与图像分析相结合以准确量化溶液状态下任何化学物质浓度的潜力。由于该模型仅依赖于样品的颜色强度而没有任何分子相互作用，它超越了比尔-朗伯定律的局限性。创建的机器学习模型还最大限度地减少了对大量专业知识和培训的需求，从而弥合了经验丰富的分析师和新手分析师之间的差距。
