How To Optimize Materials and Devices <i>via</i> Design of Experiments and Machine Learning: Demonstration Using Organic Photovoltaics

Cao, Bing; Adutwum, Lawrence A.; Oliynyk, Anton O.; Luber, Erik J.; Olsen, Brian C.; Mar, Arthur; Buriak, Jillian M.

doi:10.1021/acsnano.8b04726

Cited by 269 publications

(220 citation statements)

References 63 publications

Supporting

Mentioning

219

Contrasting

Order By: Relevance

“…is technique involved performing actual experiments (rather than numerical experiments) at specific points in the design space as determined using the selected DOE and thereafter using the data to construct the relationships. e use of experimental-based design of experiments technique is a generally accepted procedure and has been used in various disciplines such as in membranes [24], processing of food and bioproducts [25], photovoltaics [26], biotechnology [27], and analytical chemistry [28,29]. In this work, using this technique will mitigate computation costs while ensuring that all conditions specific to the injection-molding unit are captured in the analysis and, thus, leading to more accurate results of the optimized parameters.…”

Section: Introductionmentioning

confidence: 99%

Experimental-Based Optimization of Injection Molding Process Parameters for Short Product Cycle Time

Mukras

2020

Advances in Polymer Technology

View full text Add to dashboard Cite

This paper presents a framework for optimizing injection molding process parameters for minimum product cycle time subjected to constraints on the product defects. Two product defects, namely, volumetric shrinkage and warpage, as well as seven process parameters including injection speed, injection pressure, cooling time, packing pressure, mold temperature, packing time, and melt temperature, were considered. Injection molding experiments were conducted on specifically chosen test points and results were used to compute the volumetric shrinkage and warpage (at each test point). Thereafter, three relationships between the product cycle time (one relationship), the two product defects (two relationships), and the injection molding parameters were constructed using the kriging technique. An optimization problem to minimize the product cycle time (described by the first relationship) subject to constraints on the product defects (described by the latter two relationships) was then formulated. A combination set of points between the lower and upper extreme values of acceptable product defect was generated to serve as constraints for the two product defects. The optimization problem was then solved using the Fmincon function, available in the Matlab optimization toolbox. A plot of the optimization results revealed an appreciable tradeoff between the cycle time and the two product defects. To validate the optimization, an additional injection molding experiment was conducted for one of the optimization results. Results from the additional experiment showed reasonably close agreement with simulation optimization results differing in the cycle time, the warpage and volumetric shrinkage by 6.7%, 3.2%, and 8%, respectively.

show abstract

Section: Introductionmentioning

confidence: 99%

Experimental-Based Optimization of Injection Molding Process Parameters for Short Product Cycle Time

Mukras

2020

Advances in Polymer Technology

View full text Add to dashboard Cite

show abstract

“…Machine learning applied to available experimental observations and theoretical simulations could potentially generate many comprehensive models with advanced predictive capabilities. This approach has been successfully applied in several materials and molecular designs across application areas.…”

Section: Introductionmentioning

confidence: 99%

Property Prediction of Organic Donor Molecules for Photovoltaic Applications Using Extremely Randomized Trees

Paul

Furmanchuk

Liao

et al. 2019

Molecular Informatics

View full text Add to dashboard Cite

Organic solar cells are an inexpensive, flexible alternative to traditional silicon-based solar cells but disadvantaged by low power conversion efficiency due to empirical design and complex manufacturing processes. This process can be accelerated by generating a comprehensive set of potential candidates. However, this would require a laborious trial and error method of modeling all possible polymer configurations. A machine learning model has the potential to accelerate the process of screening potential donor candidates by associating structural fea-tures of the compound using molecular fingerprints with their highest occupied molecular orbital energies. In this paper, extremely randomized tree learning models are employed for the prediction of HOMO values for donor compounds, and a web application is developed. 1 The proposed models outperform neural networks trained on molecular fingerprints as well as SMILES, as well as other state-of-the-art architectures such as Chemception and Molecular Graph Convolution on two datasets of varying sizes.

show abstract

“…[43][44][45][46][47][48] For navigation of a multi-dimensional materials synthesis space, we were inspired by recent studies of machine learning regression models applied to organic electronics. 49,50 Here, we present a machine learning approach to efficiently optimize p-TCMs within the multidimensional parameter space for the CBD process. We use a strategic design of experiment (DOE) to reduce the number of required experiments.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning Optimization of p-Type Transparent Conducting Films

Wei

Gurudayal

et al. 2019

Chem. Mater.

View full text Add to dashboard Cite

P-type transparent conducting materials (p-TCMs) are important components of optoelectronic devices including solar cells, photodetectors, displays, and flexible sensors. Cu-Zn-S (CZS) thin films prepared by chemical bath deposition (CBD) can have both high transparency in the visible range (>80%) as well as excellent hole conductivity (>1000 S cm-1). However, the interplay between the deposition parameters in the CBD process (metal and sulfur precursor concentrations, temperature, pH, complexing agents, etc.) creates a multi-dimensional parameter space such that optimization for a specific application is challenging and time consuming. Here, we show that strategic design of experiment (DOE) combined with machine learning (ML) allows for efficient optimization of p-TCM performance. The approach is guided by a figure of merit (FOM) calculated from the film conductivity and optical transmission in the desired spectral range. A specific example is shown using two steps of optimization using a selected subset of 4 experimental CBD factors. The machine learning model is based on support vector regression (SVR) employing a radial basis function (RBF) as the kernel function. 10-fold cross-validation was performed to mitigate overfitting. After the first round of optimization, predicted areas in the parameter space with maximal FOMs were selected for a second round of optimization. Films with optimal FOMs were incorporated into heterojunction solar cells and transparent photodiodes. The optimization approach shown here will be generally applicable to any materials synthesis process with multiple parameters.

show abstract

How To Optimize Materials and Devices via Design of Experiments and Machine Learning: Demonstration Using Organic Photovoltaics

Cited by 269 publications

References 63 publications

Experimental-Based Optimization of Injection Molding Process Parameters for Short Product Cycle Time

Experimental-Based Optimization of Injection Molding Process Parameters for Short Product Cycle Time

Property Prediction of Organic Donor Molecules for Photovoltaic Applications Using Extremely Randomized Trees

Machine Learning Optimization of p-Type Transparent Conducting Films

Contact Info

Product

Resources

About