Spectroscopy

by

Automating the Optimization of Locally Weighted Models is a Solution

Automation of Local Regression Model Building for Spectroscopic Data, JChem2024 Pell et alA calibration model tends to improve as additional calibration samples are added to the library. If the samples reflect variation in the chemical composition, the model then expands its zone of relevance. Ultimately, as these new zones expand calibration scope, the model can degrade in performance due to nonlinearities and may require adding to the model rank, which can make the model fragile. Building local models is one answer to gain the best of both worlds, but optimizing a locally weighted model is tricky and time consuming. Automating the optimization of locally weighted models is a solution.

http://doi.org/10.1002/cem.3637

For additional information or questions, contact info@infometrix.com.

#chemometrics #LWR #spectroscopy #regression

by

Augmented Models

Picture a situation where a company has a spectrometer and model that is performing well in a quality control setting. The company wants to set up a second spectrometer for another line or another location. Normally, we need to wait until enough spectra and reference values are available to be able to build a competent model, which can take months, even a year. In Ai-Metrix, a customer can choose to jointly model a data-rich spectrometer with a data poor instrument. If Ai-Metrix sees a discrepancy in data amounts, it clicks in a new set of optimization parameters and builds an augmented model that should be available within days. The model will not be as good as a well-populated calibration, but the system would then dial itself in as more data becomes available. Because Ai-Metrix does not care how many models a user creates, the marginal cost for this series of model updates is essentially zero. And you can put the spectrometer to use right away.

by

ATC 2024 – Fundamentals of Quality Control Employing Optical Spectroscopy

ATC 2024Venue: April 15-19, 2024 Galveston Island Convention Center Presented by: Brian Rohrback, Ph.D., MBA, President, Infometrix, Inc.   Abstract: Obviously, to control quality in manufacturing, one needs to have some way of measuring the quality of the process and the product. Also critical is to optimize the action plan on how to process signal to gain the information content, to deliver the answers, and to facilitate maintenance. The field of optical spectroscopy has been critical to QC operations as a set of non-destructive technologies providing insight into the chemistry of the product. Spectrometers can deliver chemical information quickly and the on-going cost of ownership is relatively low. The quality of the information content will be a function of the analytical technology behind the instrument (NIR, Raman, UV-Vis) and analyzer calibration; the calibration task is the only one that falls to the end user on a routine basis. Maintaining best practices for spectroscopic calibration and identifying areas where the process can be streamlined is critical to preserve the value in the company’s investment in optical spectroscopy. Register at https://www.analyzertechconference.org/.
by

IFPAC 2024 – Chemometrics in the Cloud

IFPAC 2024Venue: March 3-6, 2024 Bethesda North Marriott Hotel and Conference Center Presented by: Brian Rohrback, Ph.D., MBA, President, Infometrix, Inc. Abstract: In the spirit of automation, there are cloud-based tools from both the chemometrics and the general statistics realms that can be applied to simplify the work involved in optimizing a calibration. Robust statistical techniques require some set-up of parameters, but once established for an application, they are often usable in every other instance of that application. The result is a one-pass automated means of selecting optimal samples for a calibration problem and, in turn, simplifies and automates the assignment of model rank. In the end case, this means that a spectrometer can essentially become an appliance; take it out of the box, plug it in, and enjoy. The capability exists to have a spectrometer self-tune and adapt to a specific application, then keep the spectrometer in appropriate calibration completely through closed-loop control. Automation of best practices needs to include how to match laboratory reference data to spectral data, an unbiased approach to selecting validation samples, an optimal mechanism for model construction, establishing standards for quality reports, tracking model performance over time, handling process or ingredient transitions, and much more. Register at www.IFPACglobal.org/attendee-registration.
by

PITTCON 2024 – Spectroscopy and the Intersection with Machine Learning

Pittcon 2024Venue: February 24-28, 2024 San Diego Convention Center Presented by: Brian Rohrback, Ph.D., MBA, President, Infometrix, Inc. Abstract: Application knowledge and chemometrics play a vital role in the processing of all types of multivariate data into application-specific information and has been doing so for at least 50 years. There has been a not-so-subtle shift in thinking as we integrate basic concepts and the occasional hallucination in the data mining, artificial intelligence, machine learning worlds. The target is to identify combinations of our technical tools to augment or replace tasks that consume brainpower where timely response is valued, and profits are at risk. The biggest focus of chemometrics has been in the calibration of optical spectrometers. It is worth considering the subtasks:
  1. Optimizing the instrument settings for a given application;
  2. Optimizing the method parameters – preprocessing, transformations, wavelength ranges;
  3. Handling of calibration transfer; and
  4. Optimizing models for inliers and rank in pursuit of routine processing and adjusting to changes in ingredients and unit operation.
The first two tasks are a set-once method development and the third may be generic across all applications. This paper tackles subtask 4 with a project that combined traditional approaches in statistics, database organization, pattern recognition, and chemometrics with some newer concepts tied to better understanding of data mining, neurocomputing, and machine learning. The future goal is to automate spectroscopy calibrations such that it is possible to have instrument systems tune themselves. Register at www.pittcon.org/register/.