A database solution to ease and automate processing and maintaining of fragmented data, allowing for better chemometrics, more experimentation, more frequent model updates, and optimize use of personnel.
Built on the amalgamation of experience from 15 refinery optimizations and a lot of R&D effort over the last 5 years, the whole process of Ai-Metrix is designed to take less than one hour, including transit time. Ai-Metrix uses an evaluative process that provides a database with search and matching capability and a dashboard interface allowing a user to visualize trends and comparisons of the data through a graphical output; the matching of such data being essential to calibration studies and future predictive models used in laboratory and process environments. Additionally, Ai-Metrix can provide calibration modeling and prediction capabilities in routine and automated situations with little or no human interaction. A service component is also available in Ai-Metrix, in which Infometrix staff develop, review, implement and maintain calibration services for the end user. Read about the process below or view the Ai-Metrix Brochure in pdf format by clicking on an image below.
Benefits of Ai-Metrix
- Lowers cost, speeds response time
- Scales up easily, expands easily
- Ensures a common baseline for calibration quality
- Allows comparison across locations, facilitates networking
- Enables continuity in the face of personnel changes
- A simplified approach to assembling lab values and spectra into a common database.
- Long-term management both of these data and the chemometric models generated in a browser-based, customizable, interactive set of visualizations.
- An objective and automatable procedure for selecting the samples that would be best for the calibration.
- Automated guidance on how many factors to include in the model.
- An automated mechanism for determining that a calibration requires updating and a more objective way to do those updates.
- A universal system that is completely independent of ANY manufacturer of spectroscopy gear and will support ALL legacy software.
- The result insures uniform quality models across the organization.
Chemometric Modeling
Model Request
The request is made through an Excel workbook that is prepopulated with customer-specific properties, categories, and date ranges which formats the request into a file that is immediately processed. The resulting models are automatically delivered to the desktop. The AI system picks up the request for models and sets a multi-step process in motion:
- A 3-step Robust outlier identification mechanism trims the spectra/reference pairs to eliminate samples that would degrade the model
- A 2-step PLS modeling process identifies the optimum number of factors
- A Pirouette file with all the data plus an exclusion set with only the samples and variables used to generate the model is created
- A Pirouette model that can be used for routine QC assessments, plus any supported non-Infometrix model file as specified by the user, OR the files necessary for the user to build the model in any chemometrics software is also created.
Model Quality Control
Models can be built for any property and any subset managed by the database. These data are trimmed of outliers, a PLS model is constructed, and the number of factors is set according to Infometrix best practices. The models are delivered to the customer computer within minutes. Model visualization and performance over time is available on an instrument-specific Tableau dashboard.
Base Service
The base service is designed to handle up to 20 properties per prediction, but any number of models can be created and maintained. We consider these to be preliminary models, but can serve as a baseline for any optimization that the user may do. A one-time setup cost and monthly fee applies.
Contact us now for more information, WebEx presentation of the process or a quote.
Infometrix, Inc. • 11807 N Creek Pkwy S • Suite B-111 • Bothell, WA • 98011 • tel: 425-402-1450 • info@infometrix.com
This page has the following sub pages.