Troubleshooting Calculations Transform¶
Check out the Calculations Transform documentation and the Calculations Transform Tutorial example for detailed instruction.
Q: Why is processing high-frequency data failing or very slow?
| Problem | Solution |
|---|---|
| Raw high-frequency data (e.g., 100 Hz across multiple signals for 30 days ≈ 1.3B points, ~10 GB) can exceed RAM limits and create excessive I/O. | - Use Calculations Transform to create downsampled data or usable metrics at a lower frequency (e.g., 10 Hz or 1 Hz) before feeding it into other Falkonry tools. |
Q: How does Calculations Transform impact resources?
| Problem | Solution |
|---|---|
| Calculations Transform increases resource usage. | - Requires incremental compute resources. - May reduce parallelism in model learning/evaluation. - Creates additional storage for calculated signals. |
Q: How can I use calculated signals in downstream analysis?
| Problem | Solution |
|---|---|
| Uncertainty on how to use calculated signals with other Falkonry Tools. | - Output signals can be used in Patterns, Insights, or Rules like any other signal. - Create downsampled or transformed signals that provide more use in multivariate analysis, setting thresholds, anomaly detections, etc. - Ex: "RMS signals", "Standard Deviations", "Frequencies", etc. |
Q: Why are no data points output despite successful evaluation?
| Problem | Solution |
|---|---|
| - Mismatch between input and output parameters; names/IDs must match exactly. - There is an error in the python function resulting in a failure to output data. |
- Verify that input and output parameters use the same names and IDs as in the application. - Check for typos, formatting issues, or case sensitivity mismatches. - Run test cases through your function to ensure expected functioning. - Test your function with inputs like: |