Next step
What Tensor SPC Does Differently
See how Tensor SPC preserves sensor-time structure instead of flattening it.
Continue →Fundamentals
Why Traditional SPC Struggles with Structured Data
Traditional SPC tools remain essential, but structured high-dimensional data creates monitoring challenges that ordinary single-variable or flattened multivariate approaches may not fully address.
The problem is not SPC itself
Traditional SPC is strong when a process characteristic can be monitored clearly as an individual variable, subgroup, or conventional multivariate vector. The difficulty appears when modern process data has meaningful structure that should not be discarded.
Examples of structured data include:
- multiple sensors observed over time,
- feature profiles measured through process stages,
- spatial measurements across a surface or component,
- batch trajectories with multiple variables and phases.
Vectorization can hide structure
A common approach is to flatten structured data into a long vector. This allows conventional methods to operate, but it can obscure the relationships that made the data meaningful.
| Structured process data | Flattened representation |
|---|---|
| Sensor × Time behavior is preserved. | Sensor and time relationships become positions in a long vector. |
| Localized patterns can be interpreted visually. | Localized structure may become difficult to trace back to process meaning. |
| Anomaly location can be tied to variables and time regions. | Anomaly contribution may be spread across many vector elements. |
Structured anomalies may be weak individually but strong collectively
Some process changes do not appear as a dramatic spike in one variable. Instead, they appear as a relationship change across several variables or time regions. Traditional SPC may show only a mild concern, while the structured pattern is more meaningful.
Tensor SPC is useful when the relationship pattern matters, not only the individual measurement value.
Next step
The next page explains how Tensor SPC addresses this problem by preserving multiway process structure instead of collapsing everything into a single flat representation.
Continue: What Tensor SPC does differently →Related concepts