2.6 Lab User Screen

Theo Wilson -

Lab User Screen

 

The Lab User screen is designed to allow a more detailed look at the chemical signature of the gas sample by continually building up DF Matrices It also allows the user to see both the Positive and Negative Ion Mode spectra at the same time.

The Lab User mode will not run in parallel with Process Monitoring.

 lab_user_screen.png

 


Clicking the ‘Start’ button will trigger the data collection.

 

A popup box will appear to select a filename for data storage. Either select a name from the drop down box (if available), or enter a new one. Leaving the dialogue box empty or clicking Cancel will turn data logging off.

 

All DF Matrix files can given an optional prefix which can be altered at any point during data collection by changing the text in the Prefix box.

 prefix.png

 

E.g. With “Control” in the Prefix box the DF Matrix will be saved with the filename “Control_Matrix_X”

 

The resulting Matrix file can be opened using the Lonestar DF Matrix review application or using any software package that can open “.asc” files (e.g. Excel, Open Office, Notepad).

 

Contact your Owlstone Representative for details on how the interpret the DF Matrix file.

 

Note that the Lonestar system only has limited storage space. If large data files are going to be generated then it is recommended that you either network the system or plug in an external hard drive. The default file path should then be altered in the Settings screen.

 

When running the Lonestar will scan through the DF Matrix settings used in the Settings tab and build up a complete fingerprint of the chemical sample. Once a DF Matrix is complete it will store the data to file (if file saving is enable), clear the screen, and then start collecting another dataset.

 

 

The Lab User mode uses the same Settings as the Process Monitoring mode where applicable:

 

  • The Field Intensity section determines how the DF Matrices are created and the ‘Average CV sweeps’ allows averaging to be included to improve system sensitivity
  • The Pump, Recirculation and the Heater switches all perform the same functions
  • The Comments, Analytes and Humidity boxes can be used to record additional experimental data and information
  • The Process Monitoring settings have no effect
  • The Event Settings (perform event and event length ) have no effect 

 

 lab_user_2.png

 The Lab user screen can be used to identify field intensity values (%) of interest to process monitoring

 


Review Process File

 The ‘Review Process File’ application is accessible from the File menu under the Applications menu.

 review_processfile.png

 

This application can be accessed whilst the Lonestar is in use and collecting data. To return the main control screen use the File menu to Exit.

 The Process Review screen has two main functions:

  1. Review all data collected during process monitoring
  2. Provide a means of setting up and refining alarm trigger criteria

 

The Process Review screen has a similar layout to the online process monitoring screen (Home) but has the ability to view historical data.

 

Figure_16.png

The process review screen

 

The Process Review screen has a different navigation bar, removing the options to alter the Lonestar settings and replacing them with data handling tools. The main differences are:

            A – Displays the date and time of data collection

            B – Ability to pan and zoom through time

            C – Train the system to teach it what is normal

            D – Additional weighting factors to refine the Training

 

Process Review

 To load a data file for review access the File menu and select Load File. The data file will load showing the first 600 seconds of data collected as default.

 Figure_17.png

Active process monitoring with 2 events

 

  • The top right graph shows a selectable portion of the Process History.
  • Use the zoom and scroll bars to view any data of interest.
  • All areas of the Process Review screen will update simultaneously as you pan and zoom through the data.
  • The box above the scroll bars gives the date and time that the data was collected
  • The top left graph shows the CV sweep obtained at the cursor position in the main viewing graph. Use the Rescale Ion Current bar to zoom in and out
  • The alarm reporting table shows all alarms that were triggered during the window of data being reviewed
  • The alarm level bar will also show all alarms and events that were triggered during the window of data being reviewed
  • The bottom right graph is used to display the same system information that is available in Online Process Monitoring. It is also used to set threshold levels when Training the system

  

Exporting Data

In order to be able to open the Lonestar Process Data files with spreadsheet or notepad packages the file must be ‘Exported’ to an ASCII format. This function allows the user to select the portion of data of interest before performing the export.

 export_data.png

  • Open the process file as normal
  • Zoom and scroll to the area of interest
  • Select ‘Export’ from the File menu
  • There may be a slight pause as the system processes the data
  • A dialogue box will appear where the exported filename and location can be set
  • Clicking OK will start the export process (this may take some time if the data set is large)

 

The exported file will have a “.asc” file extension. This can be opened in a variety of software packages. Note that the Lonestar Process Review screen cannot open these exported files.

Contact your Owlstone Representative for details on how the interpret the resulting exported file.

  

Training the system

Training the Lonestar is the key to allowing it to identify changes in the process that it is monitoring. The fundamental principle is that the system needs to be ‘taught’ what is normal and also ‘taught’ how to treat deviations from this normal response. Ideally the system will be trained with datasets collected during Process line sampling while it is behaving normally and then tested against examples of when the Process line has deviated from normal.

Note that the system Training is only applicable for data collected at a single, fixed RF field intensity (%).

After training the Lonestar will be able to calculate an ‘Alarm Level’ which monitors how much a single CV sweep deviates from the Averaged CV sweep that the system has been taught to consider normal. This Alarm Level will always monitor absolute deviations from normal, i.e. if a peak shrinks or increases in height then in both cases the Alarm Level will increase (it can never be less than zero).

The images below shows the steps taken in calculating the ‘Alarm Level’

  

Figure_18.png

Calculation method for deviation from normal alarm

 

Basic – Teach the system what is normal

 The first step is to train the Lonestar to show it what is considered to be a ‘normal’ spectrum.

Load a data model and zoom in on an area which is believed to have normal, steady behaviour with no features that should trigger an alarm. The amount of data required to train the system is variable but it is better to use as large a data set as possible.

This training will apply to both the positive and the negative mode at once, even though only one mode can ever be visible at one time. Therefore remember to view both positive and negatives modes whilst selecting the area to use for training. 

Once a suitable area of data has been selected, press the Train button. The software will then calculate the average CV spectrum obtained from all the data in the main data graph and display this averaged spectrum on the top left graph, as shown in red below.

  

 Figure_19.png 

Training the normal condition

  

The bottom right graph will switch to showing Ion Current Deviation. The deviation should be flat and predictable.

 The system can be re-trained on new data at any time by clicking the Train button.

In the simplest case the threshold levels can now be set by dragging them to the appropriate levels and the setup should be saved by accessing the File menu and selecting ‘Save Configuration’

  

Intermediate – View deviations from normal

If the data collected shows examples of deviation from what is regarded as normal these can be reviewed and used to improve the Training further by taking the following steps: 

  • Pan and zoom over the data set until there are examples of both normal and deviations that would be expected to trigger an alarm
  • Observe how the Ion Current Deviation varies and set the alarm levels accordingly for both the positive and the negative mode
  • Pan and zoom to any other data to confirm that the threshold levels are appropriate
  • Remember to check both positive and negative mode data

  

 Figure_20.png

Deviation monitoring

 

In the example above the system was initially trained on the first 600 seconds of data. After 700 seconds of data collection the system was exposed to a minor chemical change in the Process line and then a much more significant change after 900 seconds. These two events were used to trigger the appropriate Amber and Red alarm levels. 

Once the desired threshold levels are set the Training parameters should be saved by accessing the File menu and selecting ‘Save Configuration’. Further refinement can be achieved by altering the model weightings (Advanced users)

  

Advanced – Set up model weighting

The training can be further refined by setting up different weighting levels at different CV values. This allows the user to place additional emphasis on changes in peak heights in certain CV regions.

In the simplest case the model weightings will all be set to Unity and thus any changes at any point along the CV scale will have an equal effect on the total alarm level.

 

alarm_level_eqn.png 

 

Where:

  • IonCurrentin  = Ion Current Intensity (AU) that is considered to be normal at CV = i
  • IonCurrentaa = Ion Current Intensity (AU) of the current CV sweep at CV = i
  • IonCurrentWeight = User defined weighting scheme
  •  = Incremental step in Compensation Voltage

 

 

The graph below illustrates how the difference calculation would work with two distinct Weightings.

 

Figure_21.png 

The weighting model

  

Alarm level = (A1 x W1) + (A2 x W2)

 

The above weighting setup will produce a monitoring system that is far more sensitive to changes in the right hand peak compared to changes in the left hand peak

 

The weightings can be set by creating marker points. Double click on the top left graph and then drag the markers around the viewable graph area. To remove individual markers drag them off the graph. To remove all markers click the ‘reset’ button above the graph. 

 Figure_22.png

Changing the weighting

  

Using this weighting system,small changes in low intensity peaks can be closely monitored and certain peaks can be ignored completely if desired. This allows the system to be finely tuned to cope with both expected and unexpected changes in the Process line.

 

Note that the exact weighting factors used will be highly application specific

 

Once these weightings are set they are applied to all data in the current Process File and the ‘Ion Current Deviation’ will update accordingly and the threshold levels can be adjusted as usual.

Have more questions? Submit a request

0 Comments

Please sign in to leave a comment.
Powered by Zendesk