Choose Language Hide Translation Bar

cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Optimizing Recipe Formulation Using Machine Learning, Sequential Learning, and Design of Experiments
Wednesday, March 12, 2025
Salon 3-Rome
Machine learning (ML) methods have been widely applied to analyze design of experiments (DOE) data in such industries as chemical, mechanical, and pharmaceuticals, yet receive limited attention in the food industry, especially for recipe optimization. To address this, we explored ML and sequential learning for recipe formulation, aiming to optimize product quality. We combined DOE with ML to select optimal combinations of one, two, or three ingredients from 12  candidates, adjusting ingredient dosages based on the number of combined ingredients: 0-1 for single ingredients, 0-0.5 for pairs, and 0-0.33 for triplets. After assessing the main effects of all 12 ingredients, we narrowed the focus to five key ingredients. A full factorial design was applied to two-ingredient combinations, alongside collecting one data point at maximum dosage for each triplet. Three promising combinations were further analyzed using a space-filling design to explore the full parameter space. Subsequently, ML models were developed to predict product quality, with sequential learning guiding additional experiments to refine the model for one specific combination. This approach identified the optimal mixture with fewer than 100 lab experiments, demonstrating the efficiency of combining ML, sequential learning, and DOE in reducing experimental efforts while identifying high-performing ingredient mixtures.
A JSL Application for Accelerated Stability Modelling
Wednesday, March 12, 2025
Salon 7-Vienna
If a pharmaceutical product is likely to degrade after 12 months, how can we get an early insight into that risk? The answer is to perform an accelerated stability study that collects data over a period of a few weeks under high-stress conditions, such as elevated temperature, pH for liquids, and moisture for solids. Once a model has been selected, it can be extrapolated over time to assess the overall level of stability. Multiple impurities contributing to degradation are modelled individually and evaluated collectively to identify the highest risk factors. For moisture-driven impurities, packaging can be used to control degradation rates. A packaging model can be constructed that accounts for moisture permeation and adsorption (in both product and desiccants). A composite model (kinetic plus packaging) evaluates overall stability. This JSL application is an interesting combination of physics, chemistry, model fitting, and statistical estimation. The live demonstration covers the workflow navigation, modelling moisture vapour transmission rates, kinetic model fitting and ranking, the use of diagnostic plots and prediction profilers, and the generation of dashboards to present final results. Special emphasis is given to the methodology of model selection and the challenges that are unique to accelerated stability studies.
Understanding Positional Temperature Trends to Increase Testing Reliability Using JMP Pro
Wednesday, March 12, 2025
Salon 7-Vienna
Fire testing is one of the most critical and expensive aspects when developing intumescent products to protect steel structures from fire. Understanding the nature and performance of a furnace during testing is imperative for reliably interpreting the results from formulation development. In this case, the temperature data from sensors (thermocouples) in bespoke furnaces were utilised in JMP Pro to establish and understand positional temperature profiles while minimising test runs. The Functional Data Explorer was deployed as a dimension-reducing technique to describe temperature-time curves in terms of their principal components, enabling their positional element to be understood and compared directly. FPCA Score Plots showed clustering of positionally equivalent sensors with repeating tests, giving confidence in the reliability of consistent temperature profiles. Furthermore, FDOE simulation in combination with a 3D scatter plot gave dynamic understanding of temperature distributions at varying time intervals making it easy for chemists and managers to communicate. This approach not only resulted in significant test cost savings, but allowed for greater insight into the trends of the furnace, which would have been impossible using conventional analysis techniques. Analysis aligned well with expectations of a temperature gradient towards the back of the furnace from air movement to smoke exhaust.
Using Conditional Feature in Workflow Builder to Combine Six Smaller Workflows into One Workflow
Wednesday, March 12, 2025
Salon 5-London
In the past, comparisons of chamber configuration setups between chambers on the same mainframe or across different mainframes involved a manual method that took a day. Recently, an automated process has been developed using Workflow Builder in JMP 17. Individual workflow files were created to speed up the efficiency of the process, using an individual workflow, depending on the number of chambers being compared. The workflows took the config setup file and compared the reference chamber to the chamber(s) being queried. The process resulted in a report with graphs and tables based on the criticality of the configuration mismatches. This presentation showcases the effectiveness of Workflow Builder's new conditional feature in JMP 18. This feature provides a method for combining all comparison workflows into one workflow, with the correct section of the workflow determined by the number of chambers being compared, which is entered by the user in a conditional prompt. The relevant workflow is selected based on the prompt entry. The end result is still the same as before but now the manual selection of the workflow file has been eliminated, allowing a comparison report to be generated across multiple platform types in one or two minutes.
A Complete Workflow for Analyzing and Reporting Experimental Data
Wednesday, March 12, 2025
Salon 3-Rome
Developing and manufacturing biopharmaceuticals involve many standardized experiments and reports. Most reports contain information from the associated protocol, product quality data from a laboratory information management system (LIMS), information from lab documentation, and performance data from different devices. Currently, these data sources are processed with JMP, Excel, and other software, and then assembled in PowerPoint presentations and Word documents. I demonstrate an automation of this complete process that was developed in-house with JMP. With the help of a journal, the user is guided through an automated workflow. It uses LIMS data and lab documentation to fill in, e.g., the responses of a design of experiment (DoE) series automatically and proposed illustration options once the subject matter expert (SME) has built a model. These illustrations, chromatography traces, and information from the protocol are then assembled into a PowerPoint presentation and a Word report following a preset template structure. Generating this automated report saves a significant amount of time for the SME, allowing more time to focus on interpreting the results. Furthermore, by keeping the automation in a single software, data integrity is intact and layout and evaluation standards are followed in every report.
The Prediction of Product Satisfaction by Consumer by Leveraging Laundry Diary Data
Thursday, March 13, 2025
Ballroom Ped 3
In analyzing consumer insights for laundry products, the distinction between laundry diary data and panelist-level data is crucial for understanding user experience and detergent efficacy. Laundry diary data offers real-time insights into actual consumer behaviors and preferences, reflecting genuine usage patterns. In contrast, panelist-level data, derived from consumers’ experiences over a few weeks, can be influenced by post-rationalization, where perceptions may shift due to expectations or marketing, potentially distorting product performance evaluations. This highlights the need for relying on laundry diary data for a more accurate assessment of product effectiveness. To present these insights, we utilize JMP software for advanced statistical analysis and data visualization. Key functionalities include combining data tables for a comprehensive overview, using data cleaning tools to maintain integrity, and applying Partial Least Squares (PLS) modeling to explore variable relationships. Additionally, modeling scenarios and Monte Carlo analysis are employed to simulate consumer behaviors and predict outcomes under uncertainty. By prioritizing laundry diary data and utilizing these analytical tools, we aim to enhance product development and marketing strategies, ultimately improving consumer satisfaction and brand loyalty.
From Data to Decision: Automated SQC Reporting and Seamless MES/SQC Data Extraction
Thursday, March 13, 2025
Ballroom Ped 7
In today's fast-evolving data analytics landscape, efficient reporting tools are key to driving informed decision making. This presentation introduces the complete suite of JMP tools developed to simplify manufacturing execution system (MES) and statistical quality control (SQC) data extraction and analysis for users. Of particular interest is a JMP add-in, the result of a collaboration between Syensqo and Ippon Innovation, designed to streamline the creation of SQC PowerPoint reports, fully aligned with the company's established templates.  Key features of this innovative tool include: On-demand SQC report creation: Users can quickly generate SQC reports tailored for review and customer distribution. Interactive report review: Users can interact with the underlying data, hide or exclude rows, and regenerate reports based on these adjustments. Automated report generation: The tool automates the creation of multiple reports, notifying users via email when they are ready, thus eliminating manual processes. Seamless add-in deployment: The various add-ins are deployed and updated on multiple servers, managed through the Add-in Manager for easy access by users. We showcase the capabilities of our tools within the context of our quality management process workflow. We also share insights from our implementation journey and explore how automation is transforming reporting practices.
Optimizing Mixing Processes in Disposable Vessels for Biologics Using JMP and CFD: A Scale-up Study
Thursday, March 13, 2025
Ballroom Ped 8
Fill and finish processes for biologics include a critical mixing step to reduce non-uniformities in fluids and to eliminate concentration and pH gradients formed during freezing. Proper mixing is essential to prevent protein degradation due to mechanical stress. This study characterized fluid dynamics and mixing behavior in disposable square-cross-section vessels like Cytiva’s Levmixer. It focused on a 200L commercial vessel and a custom-designed 1L vessel. Key factors such as stirrer speed, fill volume, viscosity, impeller size, and impeller placement were analyzed to assess their impact on product quality using protein solutions as surrogates. JMP was used to design experiments and generate a predictive model for small-scale mixing. In parallel, computational fluid dynamics (CFD) simulations provided insights into fluid flow and mixing in the larger vessel. The comparison of JMP’s model with CFD simulations showed a strong correlation, with consistent predictions for mixing times and conditions across both scales. This complementary approach reduced development time and costs while ensuring product quality during scale-up from 1L to 200L, highlighting the value of combining JMP with CFD for scalable process optimization.
Adapting Design of Experiments (DOE) to Enhance Device Performance in RIE Fabrication at SCD
Thursday, March 13, 2025
Ballroom Ped 1
SCD (SemiConductor Devices), a global leader in the design and production of advanced infrared detectors and systems, is recognized for its cutting-edge technology and innovation in electro-optical solutions. As a fast-growing company, SCD operates in an environment where development times must be reduced, since being the first to present new products is critical to maintaining a competitive edge. To meet these demands, SCD adopts innovative approaches such as design of experiments (DOE) to accelerate process optimization and product development. In this work, we demonstrate the flexible application of DOE to improve the performance of a device fabricated through a two-stage Reactive Ion Etching (RIE) process. Initially, a series of experiments was planned using JMP DOE. Early insights into the process behavior indicated the need to adjust the design space. As the DOE progressed, a deeper understanding of the underlying mechanisms emerged, leading us to modify the process chemistry. This study highlights the importance of flexibility in DOE approaches, showing that real-world experimentation may require evolving the experimental plan. Ultimately, our work emphasizes that while DOE may not always yield a final predictive model, it can lead to valuable insights into process mechanisms, contributing to better decision making and process control.
Using JMP for Defectivity to Process Correlation: A Case Study
Thursday, March 13, 2025
Ballroom Ped 7
The semiconductor industry faces significant challenges in extracting and leveraging data to improve yield. One of the Defectivity workshop's missions is to locate and identify physical defects on production wafers caused by the process. These defects can directly impact the product electrical performance, highlighting the need for a better understanding of the correlation between defectivity levels (defect count per wafer) and the process involved. Highly complex, those processes are driven by multiple parameters, which means their relationship to defectivity is a major enabler for process tuning toward yield improvement and cost optimization. In this case study, we demonstrate the efficiency of JMP as a tool for data management and formatting from process in line collection through the following steps: Visualizing the initial extracted data to verify a hypothesis about the worst process tool. Manipulating and extending the data set for a more in-depth analysis of a previously identified process parameter, enabling correlation analysis. Quantifying dollar gains based on the analysis. Finally, a clear results display is highly beneficial for management when considering a potential process change and making informed decisions regarding the cost-yield balance.
Automation of Data Processing for Accelerated Product Development
Thursday, March 13, 2025
Ballroom Ped 4
During product development, reagent formulations for new sequencing platforms require multiple cycles of development, optimization, and robustness testing. The result? Significant amounts of data that need to be processed and compared using statistical analysis. These data handling tasks are often tedious, time-consuming, and prone to copy-and-paste errors. Automation of the data processing workflow represents the ideal solution to those issues. The Workflow Builder in JMP allows scientists with minimal programming experience to create their own scripts that automate otherwise repetitive data-handling workflows. In this presentation, a reuseable script is created for the analysis of an example: the results of a plate-based enzyme assay. Use of column formulas, column stacking, non-linear (exponential) fits, and combined data tables are demonstrated live in conjunction with basic functions of the Workflow Builder. The final workflow is debugged in real time and modified to make the script reuseable and robust, using simple JSL elements such as variables, and the “Current Data Table” and “Pick File” functions. The presentation culminates by showing how JMP scripts were leveraged to streamline data processing at Illumina, unlocking substantial time savings and faster insights for project development teams.
Efficient Strategies for Selecting Minimal Solute Sets in Linear Solvation Energy (LSER) Models
Thursday, March 13, 2025
Ballroom Ped 8
Linear solvation energy relationship (LSER) models are used in adsorption and chromatography to describe how molecular interactions influence solute retention or adsorption. These models relate the partitioning coefficient of a solute to various molecular properties, enabling predictions based on solute descriptors, which can be looked up or calculated via quantum chemistry. Mathematically, LSER models are expressed as linear equations, with coefficients obtained through multiple linear regression of experimental data from a set of solutes. Since obtaining data for solutes is labor-intensive, and solutes may have limitations (e.g., low solubility, high cost, or instability), selecting an optimal minimal set of solutes becomes important. This study discusses strategies for selecting a chemically diverse minimal solute set that minimizes the standard error of the model's coefficients. Monte Carlo simulations (performed in JMP via Python integration) are used to explore potential solutes, considering cases where solute descriptors span a limited range. Theoretical upper and lower bounds for the standard error are presented. Both homoscedastic and heteroscedastic LSER models are considered. Finally, the impact of interdependencies among solute descriptors on the statistical robustness of these strategies is discussed.