Automatic Evaluation and Identification of DSC Curves
April 2, 2014
The new NETZSCH DSC 214 Polyma makes DSC investigations easier and more comprehensive than ever before. It is comprised of a series of smart innovations which together addresses all aspects of a DSC investigation (see figure 1): The totally new Arenafurnace and Corona sensor in the instrument along with the novel Concavus sample pans allow for extremely fast and reproducible, high-quality measurements.
A highlight of the new software is the SmartMode user interface which greatly facilitates operation. The unique AutoEvaluation even allows for completely autonomous evaluation of the measurement – reproducibly and reliably.
The final step in a complete DSC investigation is the interpretation of the measurement results. Until now, interpreting DSC curves required both a certain level of experience and a certain expenditure of time – in order to conduct literature research for example. Thanks to the new DSC curve recognition and database system Identify, interpretation is now - for the first time in the history of thermal analysis - significantly easier and faster.
Figure 1: All aspects (360° view) of a complete DSC investigation
With Identify, known database curves and literature data are used to recognize and identify unknown DSC curves, ultimately leading to interpretation of the DSC result. The process can be applied to a totally unknown sample or to routine quality control (QC), where Identify can show a sample’s level of agreement with saved measurements for samples already deemed as acceptable. Identify is thus an extremely powerful tool for any day-to-day task such as material identification as well as for quality control and failure analysis by means of DSC. It also serves as an archiving system, since it allows direct access to the evaluated DSC curves stored in the database. As a basis, libraries with measurements and literature data from the known NETZSCH “Thermophysical Properties of Polymers” poster are supplied. Users can additionally create and manage libraries and classes of their own incorporating their own measurements and their knowledge, too. And the best for last: Identify provides results with only a single click!
Material identification is illustrated in figure 2, where Identify was executed on a non-evaluated DSC curve. With a single click, the DSC curve was autonomously evaluated by the NETZSCH Proteus® software using AutoEvaluation, and the Identify results were immediately displayed (see inset): The hit list on the left shows measurements and literature data from the database sorted according to their similarity to the unknown DSC curve. On the right, a second hit list indicates the similarity between the “unknown” and defined classes which are groups of measurements and literature data within the database (see below). In this example, the glass transition around 34°C and the melting peak around 178°C detected in the unknown DSC curve clearly allow the measured sample to be identified as the polymer PA12 belonging to the defined class “PA1x_semi-cryst.”.
Figure 2: Identify results after only one click on a DSC curve; the white line represents the unknown curve, the pink the most similar database curve
Application in Failure Analysis and Quality Control
Figure 3 depicts the use of Identify to analyze a DSC measurement on a rejected PA6 part (with broken clip). The melting peak of the defective part is at a lower temperature than that of the good parts, and an additional small peak was detected at 239°C. The calculated similarity to the user class “PA6_GF30_parts_passed” was thus relatively low (only 56.61%). It can be concluded that the bad part is not pure PA6, but most probably a blend of different PA6 types such as PA6 and PA66. This particular case is a typical failure analysis, since the rejected part was already broken. It is clear, however, that routine quality control on both raw materials and final products using the DSC 214 Polyma with the Identify software as a standard tool can minimize product failures in the field .
Figure 3: DSC measurement on a rejected PA6 part (with broken clip) analyzed by means of Identify
How does Identify work?
Identify uses an approach very similar to that employed by modern image recognition software for identifying persons or objects. This approach can be divided into three main tasks:
1) Segmentation of the DSC measurement curve: Significant caloric effects like glass transitions or exothermic- and endothermic effects must be identified and distinguished from irrelevant parts of the DSC curve. This extraordinarily challenging task is performed reliably for most of the measurements by AutoEvaluation – without any user interaction required.
2) Extraction of the properties of the effects found: Properties such as extrapolated onset temperature or peak area are evaluated automatically according to known DIN or ASTM E standards using classical features of the NETZSCH Proteus® software.
3) Recognition of the DSC curve: The unknown DSC curve is compared not only to specific database measurements and literature data but also to classes. Similarity values are computed practically in-situ using advanced mathematical algorithms. An optional selection of the temperature range allows for example for restriction to only certain DSC effects. Furthermore, algorithm types for single- or multi-component samples as well as different parameter setups are available which would take into consideration any additional information on the sample the user may have.
Of course, it is also possible to use manually evaluated DSC curves in Identify or to modify the results supplied by AutoEvaluation. Then task 1) would be carried out at least in part by the user.
Thanks to this curve recognition technology, Identify can search through hundreds of database entries and – within a single second – find the DSC measurement curves most similar to the unknown curve. Last but not least: The effect-based algorithms allow Identify to work not only with true measurement results, but also with literature data as library entries.
As shown already above, Identify can on the one hand compare the unknown curve to database entries on a one-on-one basis (“queries”); on the other hand, classification can be applied which would assign the unknown curve to certain classes (see figure 4). Such classes might be material classes (MCs) containing, for example, all measurements obtained for any PA6 polymer samples. Quality classes (QCs) would contain, for instance, only those PA6 measurements stemming from good parts which successfully passed the quality control. Some material classes like PE or PA are already included in the software, but users can additionally create their own classes which will ”learn” as each new member is added. Such classes also incorporate the user knowledge and expertise into Identify, whereas query results do not apply any such knowledge. Those depend – if AutoEvaluation was employed - only on the measurement result and in no way on the user, ensuring that the evaluations and interpretations so yielded are purely objective.
Figure 4: The Identify solution which applies both, one-on-one comparisons (“queries”) and classifications
In a nutshell, Identify …
- … is a new and unique DSC curve recognition and interpretation system providing results with a single click.
- … is useful for material identification and quality control.
- … is both easy to use and sophisticated.
- … includes a database with NETZSCH libraries for polymers as a basis as well as libraries that can be created by the user.
- … manages measurements, literature data and classes incorporating the user’s knowledge.
 G.W. Ehrenstein, G. Riedel and P. Trawiel, Thermal Analysis of Plastics: Theory and Practice, Hanser Gardner Publications, 2004
 S. Schmölzer, Kunststoffe international, vol. 10/2009, S.55-57