Blog

Our Recipe for Successful Application Development

Published 13.11.2018

The background of NIR spectroscopy is heavily rooted in laboratory analysis. The recent trend, however, has been to use this powerful technology also in industrial processes, portable analyzers and even in consumer applications. Spectroscopy-based application development might sometimes sound mysterious, which has been slowing down the utilization of spectroscopy in many potential use-cases.
Our vision is to deploy the science of spectroscopy into daily use. In this blog I’ll try to explain what the critical steps of successful application development are and how to avoid the most common mistakes when developing material sensing applications.

Your samples matter

NIR spectroscopy is broadly used with all kinds of material samples: solids, liquids and even gases. Gas and liquid measurements typically require a different kind of optical interface compared to solids or powders, which are often measured by direct reflectance measurement. Typical and the most well-known applications have been designed to measure moisture and hydrocarbons and analyze food, plastics, textiles and all kinds of raw materials (RMID), e.g. in pharmaceutical industries.
A perfect sample for reflectance measurement is homogeneous, diffusive and with a diameter suitable for spectral sensors, such as the Spectral Engines NIRONETM Sensor. Transparent, highly reflective and black materials are typically challenging, and some of them get rejected before going further in the application development process.

Luckily there are always some tricks to be used. For example, trans-reflectance probes are a useful tool when analyzing transparent samples. Extremely reflective materials are tricky as well, and black materials quite often absorb all the light making it impossible to use spectroscopy in their analysis. However, even with samples that seem black to the human eye, some are not actually “black” on a NIR range and should be tested case-by-case.

The reliability of the analysis is heavily influenced also by the number of samples. If you have too few samples or the samples are not representing the measurement range required in your application, you can’t build a reliable model.

The key to optimizing performance

We typically start our own application development process by evaluating our customer’s application idea with a simple feasibility study. The purpose of this is to give our customers answers whether their application is feasible or not and if we would be able to meet the specifications from our customer, e.g. the sensitivity of the measurement. After a theoretical estimation, we then proceed to carry out the measurements with representative samples and reliable reference values.

All of the stages of the development process are critical, as the first possible error could manifest itself in the beginning stages in the reference value. NIR spectroscopy is always used as a secondary analysis, so the accuracy of the measurement must be defined by the reliability of the reference method and sample homogeneity. Non-representative samples could lead to a situation in which we might not be sure whether the application works because the results might be unreliable.

With heterogenous samples the performance can be optimized in many ways, like averaging from different spots, grinding the samples or by using a bigger spot size. The next step of performance optimization is to select the right wavelength area. Many samples can be analyzed by using a combination, 1st, 2nd or even 3rd overtone regions from the NIR spectral range. It is good to keep in mind that the absorptions will be stronger when going towards the longer end of the range, but at the same time the analysis will be more sensitive to the surface features of your samples.

Last but not least, it is important to validate your model with new samples outside of your calibration set and also test the calibration transfer between your devices. This is a vital part of the process as this will tell you how uniform your devices are but also how sensitive your model will be to the unideal behavior of individual sensors. The ideal model includes insensitivity for heterogenous samples, variations of sensors and some changes in sensor output in harsh environments.

Scanner concept brings spectroscopy science for daily use

At Spectral Engines, we’ve dedicated our life’s work in developing a solution and a platform to make all of these steps easier. Our aim has always been to make material sensing available for everyone and to all industries. This is why we have developed the NIRONE Scanner, a completely unique and unrivalled platform for designing and developing your own material sensing solution.

We would be delighted to listen to your vision of new material sensing applications! We offer a reliable, fast and transparent application development process that is based on our expertise and learnings about chemistry, sensor technologies, measurement geometries and chemometrics. We no longer wish to keep our secret recipe for successful application development to ourselves and want to share it with you and the world. Let’s give a kick start to the next generation material sensing solutions together with a piece of science.

By Anette Pakarinen, Application Specialist, Spectral Engines

Get acquainted with the benefits of the NIRONE Scanner and contact us! We’re waiting for your application ideas.