HIGH POTENCY APIS
Figure 4 – Allocation of resources for work-up & isolation to experiments with highest IPC criticality
and additional filters applied to the merged data .
The dashboard offers an overview on the general status of the project regarding yields , project progress , criticality , PARs and the expected quality impact ( severity ) extracted , based on the criticality and probability of failure of the complete underlying data .
The interface also offers a summary of incomplete or misaligned data , as a hint to better align the data or render the process more robust . For deeper investigations , the tool offers interactive filters for in-depth evaluation of the process data in a specific context down to the process parameter or quality attribute level of detail .
Every filter widget is equipped with a statistics tab that shows the typical parameter ranges including the mean and standard deviations , as well as a histogram indicating the number of the data points supporting the statistics . Dynamic plots and filtered data tables are exportable for sharing , for further evaluation , or for reporting . Users familiar with the Python programming language can use the power of the Jupyter Lab environment to extend the functionality , or to further explore the data according to their specific needs ( Figure 5 ).
Root cause analysis based on standardised data
If an experiment turns out to be OOS , it is essential to find out why . Based on data standardisation ( the distance of the value from the mean normalised against the standard deviation ), a module for impact assessment and root cause analysis links outliers in the quality attributes ( responses ) to the corresponding outliers in the process parameters ( factors ).
The interactions of outliers between factors and responses , above a userdefined threshold , are visualised to rule out or to identify a potential root cause within the tracked process parameters . A further module compiles selected normal operating range ( NOR ) and PAR data for writing of reports . The module splits the process data into NOR and PAR experiments based on the experiment classification by the user , or by frequency of process parameter values .
NOR results are summarised at the top of the report , for reference , followed by the selected PAR results at the bottom . Statistics for the selected process parameters and a criticality plot are provided for information . If the experiment metadata contains the section numbers of the PAR report , relevant results are extracted by selection of a single filter .
The modular architecture of the tool offers the flexibility to be adapted based on task- , data- or user-specific requirements . The functionality of the different modules reflects the extent of data typically available at the different stages of a project .
For early phase projects , the extent of manually collected process parameter data is usually rather limited , while the body of the electronically tracked analytical data may already be substantial . Therefore , the tool also offers the flexibility for basic data navigation at the experiment level , in the absence of detailed process parameter data . ●
Lukas Brändli
MANAGER , PR & D
Figure 5 - Merged process & analytical data with corresponding statistic
CARBOGEN AMCIS k + 41 58 909 03 30 J lukas . braendli @ carbogen-amcis . com j www . carbogen-amcis . com
18 SPECIALITY CHEMICALS MAGAZINE ESTABLISHED 1981