Choose your reading route
eduKateSG · Why Mathematics? · Canonical Variate Analysis for Dynamic Condition Monitoring: Past-Future Correlation, State Subspaces, Residuals and Fault Detection
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 6–10
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 11–15
Check assumptions
See where design and calibration matter.Route 4 · Chapters 16–20
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 21–24
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Why mathematics is important becomes practical when a monitoring method must distinguish predictable dynamics from new behaviour. Canonical variate analysis (CVA) builds past and future data blocks, whitens them, and finds linear combinations with maximum cross-correlation. Strongly correlated pairs represent predictable dynamic state information; weakly correlated directions form residual spaces useful for fault monitoring.
For a teaching example, suppose whitened past-future cross-covariance is diag(0.8,0.3). Singular value decomposition gives canonical correlations 0.8 and 0.3, with squared shared variation 0.64 and 0.09. Retaining the first pair treats the second direction as weakly predictable. A future canonical coordinate of 1 entirely in that second direction has residual squared 1. The arithmetic is simple, while real CVA requires estimated covariance matrices, regularisation, block-length choice and validated limits.
Did You Know? CVA asks a predictive question: which combinations of the past are most correlated with combinations of the future?
This article answers one intent: The mechanism rotates whitened past and future blocks into maximally correlated pairs, separating predictable state directions from residual directions.
Canonical variate analysis uses whitened past and future blocks to find paired directions with maximum correlation. The worked diagonal example makes canonical correlations, retained state variation and future residual energy visible. Later chapters add block length, regularisation, operating modes and state interpretation. Students transfer covariance, scaling and singular directions; practitioners transfer a dynamic state-space view that distinguishes predictable structure from innovations.
Full chapter index
Chapters 1–5 · Understand the question
Chapters 6–10 · Reproduce the example
Chapters 11–15 · Check assumptions
Chapters 16–20 · Explore applications
Chapters 21–24 · Study and verify
Begin with the object being measured, because a detector name is less useful than the statistic it changes. Dynamic condition monitoring needs relationships across time, not only same-time covariance. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Past pressures predict future flows. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Static correlation can miss directionality. Build time blocks. Transfer the idea by changing one input at a time and explaining the direction before computing.
A useful audit begins by writing the claim in plain language, then identifying the observation, reference model and comparison rule. In this chapter, the central claim is: Dynamic condition monitoring needs relationships across time, not only same-time covariance. Test it first on a short sequence where every intermediate value can be inspected. Then change one condition and explain why the conclusion does or does not survive.
The mathematical value appears when a pattern is connected to one reproducible calculation. Stack selected past outputs and inputs into one vector. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Two previous samples form a compact history. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Long blocks raise dimension. Choose from process horizons. Keep raw evidence beside the transformed statistic so diagnosis remains possible.
For transfer, sketch a before-and-after case and label what is measured, what is predicted and what is inferred. The concrete case here is: Two previous samples form a compact history. That example is illustrative rather than evidence about a particular school, learner or machine. A sound explanation keeps calculation, interpretation and action in separate sentences.
Separate the mechanism from the interpretation so a number does not become unsupported certainty. Stack current or future observations to define what the past should predict. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
One-step future outputs become the target. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Future blocks are unavailable in real time until they arrive. Align monitoring carefully. Compare with a simple baseline and an alternative cause before deciding.
Reliability work needs a counterexample as well as a successful example. Here the important boundary is: Future blocks are unavailable in real time until they arrive. Ask what other mechanism could create the same observed pattern, what additional measurement would separate the explanations, and how uncertainty should be recorded. This turns a formula into a defensible decision process.
Keep timing, units and assumptions visible before applying a threshold. Use baseline means and covariance estimates for both blocks. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Whitening makes canonical correlations scale-free. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Ill-conditioned covariance destabilises the transform. Regularise transparently. Explain the key equation aloud and point to the term carrying each meaning.
A student can practise this idea without specialised equipment. Start with a tiny table, reproduce the stated operation, and predict how the output changes when one entry is doubled, delayed or removed. The practical next step is: Regularise transparently. Explain the result aloud, including the assumption that would most change the answer.
Make the example auditable by predicting direction, calculating and checking a nearby case. Estimate how past and future blocks co-vary. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
The teaching whitened matrix is diag(.8,.3). Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Finite samples create spurious correlations. Validate out of sample. Report observation, assumption, calculation and authorised action as separate layers.
A useful audit begins by writing the claim in plain language, then identifying the observation, reference model and comparison rule. In this chapter, the central claim is: Estimate how past and future blocks co-vary. Test it first on a short sequence where every intermediate value can be inspected. Then change one condition and explain why the conclusion does or does not survive.
Begin with the object being measured, because a detector name is less useful than the statistic it changes. Singular values give ordered maximum correlations between linear block combinations. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Values are .8 and .3. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
High correlation is not causation. Interpret as predictive association. Transfer the idea by changing one input at a time and explaining the direction before computing.
For transfer, sketch a before-and-after case and label what is measured, what is predicted and what is inferred. The concrete case here is: Values are .8 and .3. That example is illustrative rather than evidence about a particular school, learner or machine. A sound explanation keeps calculation, interpretation and action in separate sentences.
| Object | Value |
|---|---|
| Whitened cross-covariance | diag(0.8,0.3) |
| Canonical correlations | 0.8,0.3 |
| Squared shared variation | 0.64,0.09 |
| Retained dynamic direction | first pair |
| Future residual example | second coordinate 1 |
| Residual squared | 1 |
The mathematical value appears when a pattern is connected to one reproducible calculation. Squaring canonical correlations gives .64 and .09 in the example. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
The first pair carries stronger shared dynamics. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
It is not a universal explained-variance measure. Label it carefully. Keep raw evidence beside the transformed statistic so diagnosis remains possible.
Reliability work needs a counterexample as well as a successful example. Here the important boundary is: It is not a universal explained-variance measure. Ask what other mechanism could create the same observed pattern, what additional measurement would separate the explanations, and how uncertainty should be recorded. This turns a formula into a defensible decision process.
Separate the mechanism from the interpretation so a number does not become unsupported certainty. Retain strongly correlated canonical variates as dynamic state coordinates. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
The first pair is retained. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Dimension choice changes sensitivity. Test fault performance. Compare with a simple baseline and an alternative cause before deciding.
A student can practise this idea without specialised equipment. Start with a tiny table, reproduce the stated operation, and predict how the output changes when one entry is doubled, delayed or removed. The practical next step is: Test fault performance. Explain the result aloud, including the assumption that would most change the answer.
Keep timing, units and assumptions visible before applying a threshold. Weakly correlated directions represent less predictable variation. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
The second coordinate becomes residual. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Normal unmodelled dynamics can also live there. Calibrate baseline. Explain the key equation aloud and point to the term carrying each meaning.
A useful audit begins by writing the claim in plain language, then identifying the observation, reference model and comparison rule. In this chapter, the central claim is: Weakly correlated directions represent less predictable variation. Test it first on a short sequence where every intermediate value can be inspected. Then change one condition and explain why the conclusion does or does not survive.
Make the example auditable by predicting direction, calculating and checking a nearby case. Square the residual canonical coordinate or use a covariance-weighted form. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
A residual coordinate of 1 gives statistic 1. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
One value has no universal meaning. Compare with a validated limit. Report observation, assumption, calculation and authorised action as separate layers.
For transfer, sketch a before-and-after case and label what is measured, what is predicted and what is inferred. The concrete case here is: A residual coordinate of 1 gives statistic 1. That example is illustrative rather than evidence about a particular school, learner or machine. A sound explanation keeps calculation, interpretation and action in separate sentences.
Begin with the object being measured, because a detector name is less useful than the statistic it changes. Monitor magnitude of retained state canonical variates. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
An unusual but predictable trajectory can raise a state statistic. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Residual-only monitoring misses state shifts. Use complementary indices. Transfer the idea by changing one input at a time and explaining the direction before computing.
Reliability work needs a counterexample as well as a successful example. Here the important boundary is: Residual-only monitoring misses state shifts. Ask what other mechanism could create the same observed pattern, what additional measurement would separate the explanations, and how uncertainty should be recorded. This turns a formula into a defensible decision process.
The mathematical value appears when a pattern is connected to one reproducible calculation. Past and future horizons should cover relevant dynamics without excessive dimension. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
A slow process needs longer history. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Too long repeats information and reduces usable rows. Compare predictive validation. Keep raw evidence beside the transformed statistic so diagnosis remains possible.
A student can practise this idea without specialised equipment. Start with a tiny table, reproduce the stated operation, and predict how the output changes when one entry is doubled, delayed or removed. The practical next step is: Compare predictive validation. Explain the result aloud, including the assumption that would most change the answer.
Separate the mechanism from the interpretation so a number does not become unsupported certainty. Including known controls helps separate commanded response from faults. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Valve command predicts flow. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Misaligned inputs look like faults. Synchronise signals. Compare with a simple baseline and an alternative cause before deciding.
A useful audit begins by writing the claim in plain language, then identifying the observation, reference model and comparison rule. In this chapter, the central claim is: Including known controls helps separate commanded response from faults. Test it first on a short sequence where every intermediate value can be inspected. Then change one condition and explain why the conclusion does or does not survive.
Keep timing, units and assumptions visible before applying a threshold. Different faults disturb state and residual indices in distinct temporal patterns. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Sensor bias produces persistent canonical residuals. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Signatures overlap under noise. Build a tested library. Explain the key equation aloud and point to the term carrying each meaning.
For transfer, sketch a before-and-after case and label what is measured, what is predicted and what is inferred. The concrete case here is: Sensor bias produces persistent canonical residuals. That example is illustrative rather than evidence about a particular school, learner or machine. A sound explanation keeps calculation, interpretation and action in separate sentences.
Make the example auditable by predicting direction, calculating and checking a nearby case. Map canonical statistics back to original variables and lags. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
One future pressure channel dominates a residual. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Whitening spreads contributions. Confirm in raw units. Report observation, assumption, calculation and authorised action as separate layers.
Reliability work needs a counterexample as well as a successful example. Here the important boundary is: Whitening spreads contributions. Ask what other mechanism could create the same observed pattern, what additional measurement would separate the explanations, and how uncertainty should be recorded. This turns a formula into a defensible decision process.
Begin with the object being measured, because a detector name is less useful than the statistic it changes. Changing means or dynamics violate a fixed CVA baseline. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Seasonal operating modes rotate canonical directions. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Adaptive updates can absorb faults. Use mode-aware governance. Transfer the idea by changing one input at a time and explaining the direction before computing.
A student can practise this idea without specialised equipment. Start with a tiny table, reproduce the stated operation, and predict how the output changes when one entry is doubled, delayed or removed. The practical next step is: Use mode-aware governance. Explain the result aloud, including the assumption that would most change the answer.
The mathematical value appears when a pattern is connected to one reproducible calculation. A gap breaks multiple past and future block rows. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
One missing point affects several windows. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Silent imputation changes correlations. Track validity masks. Keep raw evidence beside the transformed statistic so diagnosis remains possible.
A useful audit begins by writing the claim in plain language, then identifying the observation, reference model and comparison rule. In this chapter, the central claim is: A gap breaks multiple past and future block rows. Test it first on a short sequence where every intermediate value can be inspected. Then change one condition and explain why the conclusion does or does not survive.
Separate the mechanism from the interpretation so a number does not become unsupported certainty. Block vectors can exceed stable sample support. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Regularised covariance stabilises whitening. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Regularisation alters canonical correlations. Validate the full pipeline. Compare with a simple baseline and an alternative cause before deciding.
For transfer, sketch a before-and-after case and label what is measured, what is predicted and what is inferred. The concrete case here is: Regularised covariance stabilises whitening. That example is illustrative rather than evidence about a particular school, learner or machine. A sound explanation keeps calculation, interpretation and action in separate sentences.
Keep timing, units and assumptions visible before applying a threshold. A residual or state signal can trigger temporal diagnosis and inspection. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
A future residual spike prompts sensor and timing checks. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
CVA does not prove fault cause. Use independent confirmation. Explain the key equation aloud and point to the term carrying each meaning.
Reliability work needs a counterexample as well as a successful example. Here the important boundary is: CVA does not prove fault cause. Ask what other mechanism could create the same observed pattern, what additional measurement would separate the explanations, and how uncertainty should be recorded. This turns a formula into a defensible decision process.
Make the example auditable by predicting direction, calculating and checking a nearby case. The topic combines correlation, matrices, singular values and sequences. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
A learner can square .8 and .3. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Strong correlation is not certainty or causation. Explain prediction. Report observation, assumption, calculation and authorised action as separate layers.
A student can practise this idea without specialised equipment. Start with a tiny table, reproduce the stated operation, and predict how the output changes when one entry is doubled, delayed or removed. The practical next step is: Explain prediction. Explain the result aloud, including the assumption that would most change the answer.
Begin with the object being measured, because a detector name is less useful than the statistic it changes. Specify blocks, inputs, scaling, covariance regularisation, retained dimension, limits and actions. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Tests include bias, delay, drift and mode change. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Unlogged alignment prevents audit. Archive transforms. Transfer the idea by changing one input at a time and explaining the direction before computing.
A useful audit begins by writing the claim in plain language, then identifying the observation, reference model and comparison rule. In this chapter, the central claim is: Specify blocks, inputs, scaling, covariance regularisation, retained dimension, limits and actions. Test it first on a short sequence where every intermediate value can be inspected. Then change one condition and explain why the conclusion does or does not survive.
The mathematical value appears when a pattern is connected to one reproducible calculation. DPCA performs PCA on lagged vectors; CVA explicitly maximises past-future correlation. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
CVA targets predictive state structure. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
It needs two blocks and stable whitening. Choose by monitoring goal. Keep raw evidence beside the transformed statistic so diagnosis remains possible.
For transfer, sketch a before-and-after case and label what is measured, what is predicted and what is inferred. The concrete case here is: CVA targets predictive state structure. That example is illustrative rather than evidence about a particular school, learner or machine. A sound explanation keeps calculation, interpretation and action in separate sentences.
Separate the mechanism from the interpretation so a number does not become unsupported certainty. Canonical variates are learned combinations, not named physical states automatically. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Rotate the second coordinate and recompute residual energy. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Interpretation requires loadings and process knowledge. Avoid invented labels. Compare with a simple baseline and an alternative cause before deciding.
Reliability work needs a counterexample as well as a successful example. Here the important boundary is: Interpretation requires loadings and process knowledge. Ask what other mechanism could create the same observed pattern, what additional measurement would separate the explanations, and how uncertainty should be recorded. This turns a formula into a defensible decision process.
Keep timing, units and assumptions visible before applying a threshold. CVA monitoring research compares dynamic schemes on benchmark processes; state-space monitoring work provides the dynamic foundation. The importance of mathematics here is that the statement becomes a quantity or logical test that another reader can reproduce.
Recent eduKate articles cover DPCA and Kalman filtering. Follow the order of operations, keep the time index and units, and check whether the direction makes sense before trusting the number.
Benchmark performance does not guarantee local performance. Validate real regimes. Explain the key equation aloud and point to the term carrying each meaning.
A student can practise this idea without specialised equipment. Start with a tiny table, reproduce the stated operation, and predict how the output changes when one entry is doubled, delayed or removed. The practical next step is: Validate real regimes. Explain the result aloud, including the assumption that would most change the answer.
The nearest published companions are multivariate EWMA monitoring, Hotelling T-squared monitoring, parity-space sensor fault detection and unknown-input observers. The eduKateSG Mathematics Learning Hub, eduKateSingapore Mathematics World, eduKate Sengkang Mathematics Hub, Bukit Timah Tutor Mathematics Hub and eduKate Punggol Mathematics Pathway provide wider learning routes. Primary sources: https://doi.org/10.3182/20140824-6-ZA-1003.02237, https://doi.org/10.1002/aic.690430922, https://doi.org/10.1016/0169-7439(95)00076-3.
Frequently asked questions
What does CVA maximise?
It finds linear combinations of past and future blocks with maximum correlation.
What are canonical correlations?
They are ordered correlations between paired canonical variates after block whitening.
What is the residual subspace?
It contains weakly predictable canonical directions not retained as dynamic state coordinates.
Does a high canonical correlation prove causation?
No. It describes predictive association within the sampled process.
How is CVA different from DPCA?
DPCA finds variance directions in lagged data, while CVA explicitly targets past-future correlation.
Extended practice laboratory
Practice case 1: Read correlations
Setup: Use .8 and .3.
Checking target: Confirm ordering.
Practice case 2: Square correlations
Setup: Compute .64 and .09.
Checking target: Interpret carefully.
Practice case 3: Retain first pair
Setup: Place second in residual space.
Checking target: State the assumption.
Practice case 4: Residual energy
Setup: Use second coordinate 1.
Checking target: Confirm 1.
Practice case 5: Change threshold
Setup: Retain both pairs.
Checking target: Observe zero residual space.
Practice case 6: Add noise
Setup: Perturb cross-covariance.
Checking target: Inspect singular values.
Practice case 7: Longer blocks
Setup: Add one past lag.
Checking target: Count dimensions.
Practice case 8: Input alignment
Setup: Shift control input.
Checking target: Observe false residual.
Practice case 9: Mode split
Setup: Fit two operating regimes.
Checking target: Compare canonical directions.
Practice case 10: Decision memo
Setup: Report predictive inconsistency.
Checking target: Avoid causal claims.