Direct Answer

The leakage current stays around 18 mA, far below the national-standard action value and even farther below the 300 mA red line — why does the model still alarm? The S-02 model uses the CUSUM cumulative-sum algorithm to detect mean drift — 18 mA looks stable, but if the baseline is 10 mA and it has been climbing steadily for the past 8 weeks, CUSUM identifies the insulation-degradation trend 4-12 weeks before the limit is crossed; it catches the "trend" rather than the "amplitude", preferring to over-alarm rather than to miss. Waiting until 300 mA to alarm may mean the insulation has already failed — trend warning is about "lead time".

A scene: a feeder's leakage was 18 mA — the red line 300 mA — far away — but it had climbed steadily from 10 mA over the past 8 weeks — CUSUM judged an insulation-degradation trend — warned 6 weeks early — the inspection found insulation aging — the trend is far earlier than the amplitude.

1. Why an 18 mA Alarm Is Surprising

The intuition of 18 mA — far below the red line: more than 15 times away from 300 mA — feeling safe; The pitfall of red-line thinking — waiting until the limit is crossed: when crossed, the insulation may already have failed — too late; The blind spot of amplitude thinking — looking only at the current value: 18 mA does not exceed — the trend invisible; The difference of trend thinking — looking at the direction of change: 10 → 18 mA climbing steadily — degradation happening; The essence of the 18 mA alarm — not an amplitude exceedance but a trend confirmation: baseline 10 mA and climbing for 8 weeks — the trend already clear.

In one sentence: an 18 mA alarm is surprising = amplitude thinking looks at "over or not" and trend thinking looks at "changing or not" — 18 mA does not exceed, but the trend is confirmed — the trend is earlier than the amplitude.

A contrast example — a feeder at 18 mA — the red line 300 mA — operations felt safe — CUSUM judged the degradation trend. The amplitude is far away and the trend is confirmed — one looks at the present and the other at the direction — the direction matters more than the present.

2. What CUSUM Is

The meaning of CUSUM — the cumulative-sum algorithm: cumulative deviation detecting mean drift — a classic statistical process control algorithm; The logic of CUSUM — not looking at a single point: accumulating the deviation of every point from the baseline — the deviation accumulates into a signal; The detection of CUSUM — mean drift: the mean changes quietly — a single point cannot show it but the accumulation can; The advantage of CUSUM — early discovery: a tiny drift accumulating continuously — alarming before the limit is crossed; The essence of CUSUM — catching the trend rather than the amplitude: the trend is the lead time and the amplitude is the present value — the trend matters more.

In one sentence: CUSUM = cumulative deviation detecting mean drift — relying not on a single point but on the accumulation — a tiny drift accumulating continuously — alarming before the limit is crossed.

A CUSUM example — a feeder's leakage rose 0.2 mA a day — a single point could not show it — the accumulation over 8 weeks made the signal clear. CUSUM catches the accumulation — a single point is noisy and the accumulation is a strong signal — the trend emerges in the accumulation.

3. How Cumulative Deviation Works

The meaning of cumulative deviation — subtracting the baseline from every point and adding up: the deviation accumulates into a signal; The appearance of drift — continuous same-direction deviation: positive deviations accumulating continuously — the mean drifting; The cancellation of noise — random fluctuations cancel positive and negative: a real drift accumulates in one direction — the noise cancels; The trigger of the alarm — the accumulation exceeds the threshold: the deviation accumulates to a certain degree — the drift confirmed; The effect of the work — identifying before the limit: a 4-12 week lead time — the maintenance window opens.

In one sentence: cumulative deviation = subtract the baseline from every point and add up — a real drift accumulates in one direction and noise cancels positive and negative — the accumulation exceeding the threshold confirms the drift.

An accumulation example — a feeder's deviations +0.2/+0.3/+0.2 mA — continuously positive — the accumulation exceeded the threshold after 8 weeks. The accumulation amplifies the tiny trend — a single point is noisy and the accumulation is a signal — continuous same direction means drift.

4. Trend Rise vs. Sudden Exceedance

Sudden exceedance — exceeding instantly: a 30 mA jump — an amplitude alarm — fast but late; Trend rise — climbing continuously: 10 → 18 mA over 8 weeks — a trend alarm — early but needing confirmation; The difference between the two — the sudden is the result and the trend is the process: the sudden has happened and the trend is happening — the process is earlier; The handling of the two — the sudden handled immediately and the trend scheduled for maintenance: the sudden is urgent and the trend is calm — different windows; The value of the trend — turning "accidents" into "maintenance": discovered early and arranged calmly — operations proactive.

The table — trend rise vs. sudden exceedance:

TypeFeatureAlarmHandling
Sudden Exceedanceinstant exceedanceamplitude alarmimmediate handling
Trend Risecontinuous climbtrend alarmscheduled maintenance

In one sentence: the sudden exceedance is the result and the trend rise is the process — the sudden has happened and the trend is happening — the trend is earlier, turning "accidents" into "maintenance".

A comparison example — Circuit A jumped to 30 mA and alarmed — handled immediately; Circuit B at 18 mA trend-alarmed — scheduled 6 weeks early. The sudden is urgent and the trend is calm — the trend gives the window — scheduled maintenance is better than firefighting.

5. The Arrhenius Equation: The Theoretical Basis of Insulation Degradation

The meaning of Arrhenius — the temperature-reaction-rate equation: temperature rising and aging accelerating — an exponential relationship; The relationship with leakage — insulation aging accelerating: each temperature rise makes the degradation rate rise exponentially — the leakage grows with it; The meaning of the exponent — leakage is not linear: early slow and later accelerated — the later, the faster; The meaning of prediction — extrapolating the trend: based on the early exponential trend — extrapolating the failure time — warning early; The landing of the theory — the trend has a basis: CUSUM catches the trend and Arrhenius explains it — the mechanism closed loop.

In one sentence: Arrhenius = temperature rising and aging accelerating (exponential) — leakage growing exponentially — early slow and later accelerated — extrapolating the trend and warning early.

An Arrhenius example — an insulation temperature rose 10℃ — the degradation rate doubled — the leakage climb accelerated. Arrhenius explains why the leakage climbs faster and faster — the exponential trend can be extrapolated — identifying early is what brings the lead time.

6. Where the 4-12 Week Lead Time Comes From

The source of the lead time — trend extrapolation: extrapolating the current climb rate — the time to the red line — the lead time; The difference of 4-12 weeks — different degradation speeds: a slow degradation 12 weeks and a fast one 4 weeks — different windows; The precision of the lead time — an accurate baseline is the key: the baseline drifting and the extrapolation inaccurate — baseline calibration; The meaning of the lead time — the maintenance window: scheduled maintenance and spare-part preparation — arranged calmly; The value of the window — turning emergencies into plans: not firefighting but scheduling — operations proactive.

In one sentence: the 4-12 week lead time = the time to the red line obtained by extrapolating the trend — a slow degradation 12 weeks and a fast one 4 weeks — the maintenance window, arranged calmly.

The table — windows by degradation speed:

Degradation SpeedLead TimeHandling
Slowabout 12 weeksregular scheduled maintenance
Mediumabout 8 weeksplanned maintenance
Fastabout 4 weeksurgent scheduling

A window example — a feeder degraded slowly — extrapolated to the red line in 10 weeks — the scheduled maintenance completed. The lead time is the maintenance window — scheduled maintenance and spare parts in place — turning emergencies into plans.

7. Turning the Lead Time into a Maintenance Schedule Window

The meaning of the schedule window — turning the lead time into a plan: when to maintain and when to prepare — scheduling; The basis of the scheduling — the length of the lead time: 12 weeks for regular maintenance and 4 weeks for urgent maintenance — graded scheduling; The content of the scheduling — maintenance preparation: people, spare parts and power-outage plans — arranged early; The verification of the scheduling — the baseline updated after maintenance: re-baselined after the repair — model calibration; The value of the scheduling — maintenance planned: not passive response but proactive scheduling — operations controllable.

In one sentence: turning the lead time into scheduling = graded scheduling by the lead time — 12 weeks regular and 4 weeks urgent — people, spare parts and power-outage plans arranged early.

A scheduling example — a feeder had an 8-week lead time — scheduled for maintenance in week 6 — the baseline updated after the repair. The lead time becomes a plan — scheduled maintenance, spare parts in place and the outage arranged — maintenance is not flustered.

8. Preferring to Over-Alarm Rather than to Miss

The cost of over-alarming — a false alarm: the maintenance found no big problem — the cost limited; The price of missing — insulation failure: an accident, a shutdown and losses — the price enormous; The logic of the trade-off — over-alarming is lost work and missing is an accident: lost work bearable and an accident unbearable — leaning toward over-alarming; The design of the model — sensitivity first: the CUSUM threshold on the low side — preferring to alarm early; The essence of the trade-off — safety first: the warning model serves safety — over-alarming beats missing.

In one sentence: preferring to over-alarm rather than to miss = over-alarming is lost work and missing is an accident — lost work bearable and an accident unbearable — safety first.

A trade-off example — a feeder alarmed 2 weeks early — the maintenance confirmed slight aging — a false alarm but reassuring. Over-alarming is a cost and missing is a risk — safety first — a warning is better early than late.

9. Back to FEXLINK: How to Use the CUSUM Trend Warning

FEXLINK's Tiangyan Engine S-02 residual-current trend-drift model: using the CUSUM cumulative sum to detect mean drift — 18 mA looks stable, but the baseline 10 mA keeps climbing — the insulation-degradation trend identified 4-12 weeks before the limit — catching the trend rather than the amplitude. The trend-warning checklist: ① the algorithm — CUSUM cumulative deviation: a single point cannot show it and the accumulation can → ② trend first — catch "changing or not" rather than "over or not": the trend is the lead time → ③ the theoretical basis — the Arrhenius exponential aging: the leakage climbs faster and faster and the trend can be extrapolated → ④ the lead time — the 4-12 week window: a slow degradation 12 weeks and a fast one 4 weeks → ⑤ scheduled maintenance — turning the lead time into a plan: people, spare parts and power-outage plans arranged early → ⑥ safety first — preferring to over-alarm rather than to miss: lost work bearable and an accident unbearable — trend warning, turning "accidents" into "maintenance".

One summary: the CUSUM trend warning = cumulative deviation detecting mean drift — 18 mA looks stable, but the trend is confirmed — warning 4-12 weeks before the limit — catching the trend rather than the amplitude. Amplitude thinking waits for the exceedance and trend thinking has the lead time — the trend is the process and the accident is the result — trend warning, turning "accidents" into "maintenance".

One reminder: when you see leakage at 18 mA, do not just say "far below the red line" — ask first: what is the baseline? Has it been climbing over the past 8 weeks? — if the trend is there, 18 mA is also a warning signal. A low amplitude does not mean safety — the trend is the key — CUSUM catches the trend, giving you a 4-12 week window before the limit.

Scope and Limitations

  • The conclusions in this article are intended for LV distribution and electrical-safety sites. Standard numbers (GB 13955) refer to their currently effective editions, and product parameters follow the latest FEXLINK selection handbook and product knowledge base.
  • Quantitative criteria (protection zone, discharge and residual voltage, earth resistance, thresholds and grading) hold under the preconditions stated in this article. Where the site voltage level, earthing arrangement, environmental category or load profile differs, the figures must be recalculated. Specific configurations involving Tianyan engine S-02 residual-current trend drift model should be verified against actual circuit parameters.
  • Specific product models and parameters are subject to the latest selection handbook and product knowledge base; this article does not constitute selection or procurement advice.
  • Standard numbers cited here refer to the currently effective editions; when a standard is revised, the related quantitative conclusions must be re-checked.
  • Deployment and implementation for specific projects must be carried out by qualified engineers based on site survey and design drawings.

Frequently Asked Questions

CUSUM Drift Detection

The leakage current stays around 18 mA