When to Use Each Seeq App/Extension
How do I do ___ in Seeq?
First, try asking the AI General agent in Seeq.
The most common analytics functionality and calculations are built into Workbench as point-&-click Tools and default Views. See the Skill Explorer guide.
But if you want to customize your analytics, the table below can help you distinguish when you should use various tools, features, and extensibility options in Seeq.
(“X” means the given column feature can do the task in that row. See notes by the “X” for relevant links and more info.)
| Formula | User-Defined Function (UDF) in Formula | Data Lab | External ML – Sagemaker, Azure ML, etc. |
What is it? What does it do? | A place to write equations using simple math and a library of functions for industrial data analytics | A way to make a Formula “macro” (combine or package multiple Formulas or external calculations into a new Formula function)
| A Python scripting environment for easy integration of scripts and customization in Seeq. | External (non-Seeq) platforms for training and using machine learning, AI, and other advanced algorithms |
Time-series: trend, aggregations, metrics, stats, regression models, limit calculations, detect deviations (incl. across assets), statistical process control. | X – better choice: •Low-code •Results auto-update •(Better choice: no-code workbench tools if they meet your needs) | X – good choice: •Low-code, requires set-up •Results auto-update •UDFs are “formula macros” combining complex formula | X •Recommended if functionality does not exist in Formula •For auto-updating: make certain scripts available as a UDF in Formula using the Add-on Calculation Connector (ACC), or use scheduled notebooks •Requires coding
|
|
Non-time-series calculations/statistics for given time periods or conditions; e.g., average, standard deviation, delta, etc. (full list: see Signal Value Statistics page in the Formula Documentation area) | X – better choice: •Low-code •Results auto-update •(Workbench Tools are no-code ways to use common Formulas) |
|
|
|
Non-time-series math and other eqns not included in the above. e.g. ANOVA, Ordinary Differential Eqns |
|
| X | X |
Automatic updates when new data are recorded | X | X | X •Make certain scripts available as a UDF in Formula using the Add-on Calculation Connector (ACC) •Or set update schedule with scheduled notebooks (spy.jobs.schedule) |
|
See/use data in Workbench | X | X | X (see notes in cell just above) | X For models that Seeq: •Serves (hosts): ACC •Does not serve (host): run model on a schedule, store result in a datasource that is connected to Seeq Or, use SPy (Seeq python library) for data access and/or operationalization back to Seeq. |
Make a “macro”/ combine or package multiple functions/ code into a new function |
| X Compiling complex Formulas into a function | X Combine Formula and python, etc. |
|
Admin access needed? |
| •Anyone can make a UDF for themselves. •Need admin access to make the UDF available to everyone on server. | •Depends on the scope of the analytics done: may need admin access to make results available to everyone on a server •ACC requires Admin access |
|
Pre-existing Python, R, etc. algorithms |
|
| X |
|
Number of outputs | 1 | 1 | 1 or more (If you use ACC, can only output 1 since it’s in Formula) | 1 or more |
Build asset structures | X Single-level Asset Groups can be built in the Data Tab --> Asset Groups section |
| X Build and apply analytics across multi-level asset structures |
|
Custom recursive or iterative calculations (e.g. goal seek, equation of state) |
|
| X | X |
Uses data structures or types that don’t exist in Seeq |
|
| X | X |
Adv. Models: machine learning (ML & AI), neural net, etc. |
|
| X | X |
Templatize analysis, apply across assets | X Some analyses can be applied across assets if an Asset Tree exists |
| X |
|
Custom Visualization (beyond those provided in Workbench) |
|
| X
|
|
Data science applications involving: •high performance and distributed compute management •(ML Ops) management and versioning of datasets, models, and source code •modeling of hundreds of calculated features and years of data |
|
|
| X These platforms can serve (host) the models, and have the results come back into Seeq upon request Add-on Calc Connector Or, use SPy (Seeq python library) for data access and/or operationalization back to Seeq.
|
ML Model training |
|
|
| X Best for epoch-based training or large training times (many assets) |
How to access | (Homescreen --> “New” button) --> Workbench --> Tools tab --> Formula | Recommended: use the User Defined Functions Formula Editor Add-on. Or see here. | (Homescreen --> “New” button) --> Data Lab Project |
|