Upgrading with Data Quality Guidance

Introduction

Users of any laser diffraction system, including the Mastersizer 3000 and Mastersizer 3000E, know that there are numerous factors which contribute to the quality of the data achieved. Without suitable attention paid to controlling each of these factors, data quality issues can arise which can impact the accuracy and reliability of the measured particle size distribution (PSD).

Receiving training in best practice for method development and setting up measurements in particle size analysis by laser diffraction is certainly important. Malvern Panalytical provide user training and educational resources to Mastersizer users with this in mind. Reinforcement of these learnings and provision of on-going support is also extremely valuable in allowing users to establish confidence in working with PSD data and spotting their own data quality issues. 

However, as with any training process, its success relies on the learnings being retained by instrument operators and educational resources to have been engaged with. Similarly, turnover in trained staff can lead to an overall reduction in skill levels and knowledge within organisations.

So, why not build best practice and on-going support into particle size analysis software?  

Malvern Panalytical have now done just that for the Mastersizer 3000 and Mastersizer 3000E Extended with the new software add-on, Data Quality Guidance. 

When Data Quality Guidance is purchased, a licence key is provided to unlock new functionality to support users to achieve the best quality data. In this technical note, we summarise this new functionality with examples relevant to version 4.1 of the Mastersizer 3000 Full/Extended software.

What does Data Quality Guidance improve?

A summary of improvements made by Data Quality Guidance to the Mastersizer 3000 Full/Extended software within version 4.1 is presented in Table 1.

The key changes when a Data Quality Guidance licence are present are:

• Introduction of new Data Quality tab in the Measurement Manager window (1 in Figure 1)
• Improvements to background data quality reporting (2 in Figure 1)
• Providing causes and solutions statements for data quality reporting (3 in Figure 1)

Version 4.1 of the Mastersizer 3000 Full/Extended software also introduces data quality features for all users – irrespective of whether they have a Data Quality Guidance licence. These changes improve ease of use of the Mastersizer 3000 Full/Extended software for everyone, but they are particularly useful when you have a Data Quality Guidance licence. These features are:

• Changes to post measurement data quality reporting (4 in Figure 1)
• Manual %RSD checks (5 in Figure 1)

Figure 1: Key changes introduced by Data Quality Guidance – measurement manager window (left) next to main workspace window

1) New Data Quality tab within the Measurement Manager window

Data Quality Guidance introduces a new tab at the right-hand side of the Measurement Manager window. This new tab is the location where all data quality information is provided during the measurement workflow.

This tab displays data quality information for the following:

• Background data – information populates within this section upon completion of the red-light background measurement. If the background data quality is suitable, it states ‘Good’. Otherwise, the data quality issue will be flagged.
• Sample measurement – information populates within this section for each individual measurement upon completion of the red-light measurement when a data quality issue is identified. If no data quality issues are identified, no information will be populated. It is possible for multiple data quality issues to be flagged at the same time. 
• Dataset variability analysis – information populates within this section upon completion of a set of measurements. The %RSD for the dataset is calculated and assessed against ISO criteria (if more than 5 measurements performed) and USP criteria (if more than 6 measurement performed). 

2) Improvements to background data quality reporting

Without Data Quality Guidance, assessments of background data quality are limited to assessments of scattering signals at select detectors. A poor background data warning, possible window or dispersant contamination, is displayed if:

• The scattering signal on detector 15 is greater than the signal detector 1, or
• The scattering signal on detector 1 is greater than 200 units, or
• The scattering signal on detector 20 is greater than 20 units.

Data Quality Guidance introduces the assessment of background data with machine learning algorithms. Dynamic and static data assessment is performed:

• Dynamic data assessment: Variation in the signal on each detector is monitored in real time over the duration of the background measurement. If any detector exceeds a certain threshold, a data quality warning is triggered. Classical machine learning also identifies bad backgrounds and the most likely cause of the poor background e.g. bubbles, contamination, thermal instability.
• Static data assessment: Machine learning algorithms assess the shape of the background and the size of peaks in the background to identify problems caused by dirt on the cell windows, high level contamination from a previous sample or misalignment. 

Possible causes to each issue are listed (ordered from most to least likely) and easy to follow steps to resolve the issue are provided.

Example warnings provided for background measurements are shown in Figure 2.

Figure 2: Example warnings for background measurements

3) Providing causes and solutions statements for all data quality reporting 

Without Data Quality Guidance, limited guidance is provided with respect to identifying possible causes to data quality issues and suggesting solutions to these issues.

Data Quality Guidance improves on this by indicating possible causes to each issue (ordered from most to least likely) as soon as they are identified and providing easy to follow steps to resolve the issue.

Example warnings provided for sample measurements and the associated statements for cause/solution are shown in Figure 3. 

Figure 3: Example of warnings for sample measurements

4) Post measurement data quality reporting 

The appearance of the post measurement, record specific Data Quality reports has been altered slightly to allow for implementation of Data Quality Guidance. These alterations will be visible irrespective of whether you have a Data Quality Guidance licence or not.

Comparing Figures 4 (pre-version 4.1 Full/Extended) and Figure 5 (version 4.1 Full/Extended, no Data Quality Guidance), the key changes observed are:

• Version 4.1, by default, separates out the data quality issue being flagged from the cause and solution information. The solution information is also more specific – for example, when a poor data fit is achieved, the user is guided to review the data fit, refractive and absorption indices. 
• An additional visual indicator is provided to show if the sample data quality is good or subject to a warning. 

Figure 4: Post measurement Data Quality reports (pre-version 4.1 Full/Extended)

Figure 5: Post measurement Data Quality reports (version 4.1 Full/Extended without Data Quality Guidance)

With Data Quality Guidance, additional background data quality information is also reported to give a comprehensive overview of data quality information for each record selected – see Figure 6. Issue, cause, and solution information is provided. Separate visual indicators are provided for background and sample measurements, along with the overall data quality assessment. 

Figure 6: Post measurement Data Quality reports (version 4.1 Full/Extended with Data Quality Guidance)

5) Manual %RSD checks

Version 4.1 of the Mastersizer 3000 Full/Extended software introduces the ability to assess data variability using manual %RSD limits – irrespective of whether you have a Data Quality Guidance licence. This complements the existing functionality to assess against ISO and USP criteria. 

This is achieved through the addition of the ‘Manual %RSD tab’. The maximum %RSD limits for three percentiles can be specified – the default percentiles are Dv10, Dv50 and Dv90 as shown in Figure 7. 

The software will indicate if the sample %RSDs are below the limits (a pass as in Figure 7) or exceed the limits (a fail as in Figure 8). 

These percentiles can be changed to others besides Dv10, Dv50 and Dv90. For example, Figure 9 shows a case where Dv20, Dv50 and Dv80 have been evaluated. 

Figure 7: Manual %RSD check – passing result for Dv10, Dv50 and Dv90

Figure 8: Manual %RSD check – failing result for Dv10, Dv50 and Dv90

Figure 9: Manual %RSD check – alternative percentiles assessed