12th of December 2022
The 2022.R2.0 release contains the following new functionality and improvements:
Machine Learning Hub
TrendMiner’s vision to democratize analytics goes beyond empowering domain experts with self-service tools, to also include closer collaboration between a variety of experts to solve problems. Some of the most complex problems require the introduction of data scientists who bring specialized techniques to the table that allow companies to squeeze the deepest insights out of the available data. Think about advanced statistics and machine learning models.
After a very successful trial program with notebooks, we are proud to announce the full launch of a new industrial Machine Learning Hub (MLHub) for timeseries data, extending analytic and machine learning capabilities of the production client of TrendMiner. With MLHub, (Citizen) Data Scientists can access TrendMiner data (both raw as well as pre-processed and contextualized data in TrendMiner views) and validate hypotheses; and/or create/train/deploy machine learning models in the new Notebook environment, which other TrendMiner users can then apply through machine learning model tags in TrendHub or visualize in DashHub.
This product launch includes the following:
- Expert licenses now unlock the entirely new Machine Learning Hub, hosting advanced analytical tools and notebook technology
- A switch from Zeppelin (trial-program) to Jupyter Notebook ecosystem to have shorter update cycles and better support from the open source community
- Kernel isolation and resource management for improved stability and performance
- Enhanced security layer, allowing well-defined access control and role-based access to data sources for the entire MLHub functionality
- Create notebook output objects for easy visualization
- Machine learning model tags, based on PMML models
Additional New features
- Multi-scatter plot for Context views: view correlations and distribution of your context data fields to gain insights from events.
- Text Tile: You can now add text to dashboards, making it easier to share the context behind your analysis or tell a compelling story around your data
- Support for digital tags in calculations on search results: get insight into which phase/process step/Batch ID was active at the start or end of the search result.
- Restore last session: Introducing autosaving in the background, to allow the reopening of the last state of view you worked on (TrendHub, ContextHub, DashHub) after the browser closes abruptly.
- 1-click option to add an item to a new or existing dashboard making it easier to extend your existing dashboards with your latest analysis, or directly create a new dashboard.
- 1-click option to share views and dashboard making it easier to share your work and collaborate with your colleagues.
- Time navigation bar in Gantt view to easily manipulate the time span you wish to analyze.
- Search results indications: Search results are displayed at the bottom of the context bar.
- Advanced Similarity Search: The advanced option allows you to search a pattern of one tag in the history of another tag.
- Context Item Search – Layering improvements: Select and remember which attribute to visualize when the search result is linked to an asset.
- Filtering on empty/nonempty values in ContextHub to filter not only based on specific values, but also on empty or non-empty values to, for example, Identify missing information.
- DashHub Trend tile includes filtered out periods: For a trend plot, filters are shown visually in grey on the tile, while scatterplots and histograms will omit the data.
- Improved information messages with more options and possibilities to keep the message visible by hovering over or to click away a message immediately.
- Extended connectivity to AspenTech InfoPlus 21: Next to Aspen SqlPlus queries, connectivity via stored procedures is now also supported.
- Jupyter notebooks can be imported into the TrendMiner MLHub notebook environment
- Notebook outputs can now be published in specific pipeline-work-items within the work organizer for subsequent use as (interactive) visualizations in DashHub. Compared to previous Notebook tiles, these pipeline-objects are resource-scarce, decoupled from the notebook itself to avoid collisions and kernel-isolated.
- Various Work Organizer UI improvements including visual indications of folder paths, easier access to actions, etc.
- Windows Server 2022 is a supported operating system to host Plant Integrations.
- Support for self-signed certificates is added on triggering monitoring webhooks.
- Fixed an issue where modifications to or deletions of digital states in the OSIsoft PI digital state mapping table resulted in indexing errors. Digital states are currently not resynced. Please reach out to support in case digital states have been modified in OSIsoft PI.
- We do not run automatic backup for clean installations anymore.
- Create backup was failing (the fix is included as well in R1.0-18).
- The cross-asset value-based search is not available in TrendHub Next Generation.
- Scooter values may change when zooming in after adding the scooter. This is due to the fact that scooters currently interpolate the data points on the chart.
- When calculated tags are deleted, they cannot be created with the same name again.
- After renaming a calculated tag linked to a context item, the component of the context item is not updated correctly.
- The jupyter notebook files and output pipeline are not backed up automatically by TrendMiner in this release. You can recover them from your VM snapshots or upload them via the ‘import’ function.
- When sharing notebooks with other users via the Work Organizer, they are shared in view only mode. Modifications by users other than the notebook owner will not be saved.
- The action displayed on the right-side panel of ConfigHub when a service is restarted, is no longer visible.
For participants in the Data-Scientist-In-The-Loop trial program for notebooks: starting 2022.R2.0, the new Jupyter Notebook environment will become part of our expert user toolset. If you have notebooks created on the Zeppelin technology stack, they will have to be converted into Jupyter Notebooks. Export the notebooks that need to be kept before starting the upgrade and use the script that can be provided by your Customer Success Manager to convert them. After the upgrade, they can be imported via the new notebook import functionality.
Notebooks are now linked to a single user (the creator) and are part of the TrendMiner Work Organizer, instead of previous availability in a central notebook-only shared repository. As a result, sharing is no longer implicitly achieved (because of the shared repository), yet notebooks can be (explicitly) shared with other users with the help of the Work Organizer, be it in view only mode for now. These sharing preferences will need to be set manually. Please keep this in mind when importing the notebooks.
After the upgrade, existing notebook tiles based on Zeppelin Notebooks will no longer work. They will need to be manually removed from the dashboards or replaced with new tiles. This is necessary as the new notebook output tiles are directly ingesting resource scarce pipeline objects rather than selected cells in a notebook that required re-running the entire notebook. These objects persist independent of the original notebooks and allow a very user-friendly configuration.
Multi-scatter plot for context views
With the Multi-scatter plot you can plot context data fields on a scatter chart to gain insights of correlations and distributions regarding events.
The “Scatter” tab in ContextHub can be found top of the display next to “Table” and “Gantt” visualization options.
Scatter configuration can be accessed by clicking the “Settings” button - , located top right.
There are 4 chart options available:
- Colored points – data points are displayed in a color scale from blue to orange, representing the transition from oldest to newest data points.
- Correlation – additional tiles are displayed that represent the correlation coefficient together with its equation value for a specific scatter plot.
- Grid lines – grid lines appear, making it easier to read values on the charts.
- Histograms – show/hide histograms.
Listed below the layout configuration options, are all available fields that can be plotted on the chart.
Navigation between multi scatter and single scatter mode is done by using the navigation on the top.
The level that you are currently in is highlighted in blue. If you hover over a disabled navigation button a tooltip appears with additional information about why the button is disabled.
When “Inspect” is enabled you can hover over a histogram to show the range, or hover over a point (or cluster of points) to show details about them. If you click on a point/cluster of points the tooltip will become sticky. To close the tooltip– click on the “x” button located on the top right on the tooltip.
For two dropdown fields the scatter chart will appear as a bubble chart – the bigger the point on the chart, the more data points it contains.
If we inspect the two points on the chart – the one on the left side has one point:
The one on the right has five points:
Note: “Colored points” are not supported for the bubble chart.
Dashboard - Text tile
The text tile provides the opportunity to add more information, add data to a dashboard and unlocks the capability to drive storytelling and journaling with your dashboards.
The new text tile can be found in the tile list when creating a new tile.
All options are present in the tile itself for easier access and a quicker way of working.
By default, a basic set of options is displayed. Clicking on the "Show more" button reveals even more options. Which can also be hidden using the "Show less" button.[Nex text tile - Advanced options]
Restore Last Session
In the previous TrendHub we could work with the URL links to retrieve "lost" sessions due to a crashed browser, lost connections or being kicked out of a session.
With this feature, an autosave action is operated, meaning that when something unexpected happens and the browser freezes, an option appears in the next session in the form of a toast message, where the " lost view " can be restored.
This option also works in ContextHub and DashHub.
Extended AspenTech InfoPlus 21 connectivity
Exiting connectivity of TrendMiner with InfoPlus 21 was based on an approach involving Aspen SqlPlus queries, as of now, connectivity with AspenTech InfoPlus 21 historian can be established with a stored procedure.
To enable the stored procedure-based approach, use the webconfig property “ip21.history.storedprocedure“. The flag is set to “false” per default and must be set to “true” to enable connectivity via a stored procedure. The SQL code within the stored procedure must be prepared by the historian administrator, who knows the data structure and what data must be obtained.
Search Calculations – Support for Digital Tags
In previous releases only analog tags could be selected as calculation on top of search results. This release enables you to add digital tags as a calculation too. The supported operators for digital tags are the "start"- and "end"-value which gives more insight in which state the search conditions are starting or ending. E.g. Is a valve in the open or closed state, return the batchID or recipestep for a particular search result.
This added calculation can then also be leveraged in further actions like sorting per state, and including the state when exporting as with any other calculation.