Assessing Product Quality or Direction for Spectral or Chromatographic Data

Video using JMP Pro18 was posted in May 2025.

Are you responsible for making decisions about chemical processes or products based on large amounts of messy data? Is your data in multiple formats received from samples or from inline or offline laboratory equipment? Do you need to relate measurements made on a chemical system to a state or property of the system? Are you a chemical analyst on a team that needs to assure, improve or match the quality of consumer products or food products based on chemical attributes? Are you asked to make sense of and draw conclusions from spectra for mixtures?

Tools like Model Driven Multivariate Control Charts, PLS, FDE, and Gen Reg are useful to assess quality and build predictive models.

In this session we use several case studies, including aase study for near-infrared transmission analysis of gluten and starch powder mixtures, to show how to monitor a process online to deploy process analytic technology (PAT):

• Import multiple file types.
• Clean and pre-process the data interactively in a way that can be applied to new data.
• Explore the data to uncover possible patterns.
• Build, interpret, compare and select predictive models to meet your product or process goals.

Suggested prerequisites:

• Familiarity with spectral data inputs and responses.
• Some experience building predictive models                                 

Questions ansered by @Bill_Worley and @yasmine_hajar at the live webinar:

Q: What does "out of control " mean in terms of chromatographic data? Does it mean that the baseline is not good?

A: It is first step to check if any of the samples is far off the average of the rest, for example due to contamination or anything of the like.

Q: How is using JMP FDE for chromatography better than using the usual peak integration method?

A: FDE allows you to analyze the whole chromatogram rather than just few peaks. This is especially useful when there is subtle variation, process drift or misalignment of the peaks.

Q: How do you decide what data clean up needed in JMP FDE platform?

A: It depends. Maybe you see there is baseline shift between your samples, or you want to align for example the highest peak. Or, maybe you want to look at a narrower range of the spectrum.

Q: What was the name of that best software recommended for image data?

A: The Deep Torch Learning for JMP Pro Add-In  is available to use with JMP Pro.  A blog has a video on using the add-in.

Related Resources:

• Discovery Summit 2025 Presentation: How to Decode Rainbows: The Best (and Least Confusing) Ways to Analyze Spectral Data
• More about Marianne Toft (Novonesis), who is mentioned in this video, and her Discovery Summit presentations