Current Sensor Machine Monitoring: Thresholds, RFP and FAT/SAT in Thailand

“Many of our machines are old and we cannot touch their PLCs. We have heard that CT clamps can be retrofitted cheaply, but will they really give us a utilisation rate?” We often hear this from production engineering and maintenance teams at Japanese-owned factories in Thailand. Here is the short answer: in a current sensor machine monitoring implementation, what the current tells you is that “electricity is flowing”, not that “the machine is producing”. The real deliverable is not the sensor but the definition of states such as “stopped, idle and machining” (the state model) and the threshold for each machine. CTs can be retrofitted cheaply, but what decides the success of the project is threshold design and acceptance testing. And what you get from current is mainly operating time; production counts and good-part counts need separate data.

All amounts, machine counts, frequencies, headcounts, power consumption figures and payback years in this article, except the average electricity tariff, are original estimates and assumptions (placeholder values) created for this article, based on the model factory described later. They are neither industry averages nor survey results. In particular, the prices of sensors, gateways, installation work and system integration are placeholder values, because no publicly available primary information could be found. Please read them as a “calculation template” and replace them with your own measured figures and quotations.

Why Current Sensor Machine Monitoring Implementation Is Now an Issue for Thai Factories

The momentum of investment into Thailand continues. According to an announcement published by the Thailand Board of Investment (BOI) on 23 July 2026, investment promotion applications in the first half of 2026 numbered 1,299 and were worth USD 43.6 billion (about THB 1.47 trillion), up 37% year on year. However, most of the application value is in the digital sector (such as AI data centres), so this figure does not in itself represent growth in manufacturing capital investment.

As a figure closer to manufacturing, the same announcement reportedly showed 132 applications worth USD 507.6 million in the first half of 2026 under the “Smart and Sustainable Industry” measure, which covers machinery upgrades and the introduction of automation. This can be read as a sign that investment in upgrading existing factories continues on a certain scale in Thailand.

On the production side, according to Trading Economics (30 September 2026), reporting an announcement by Thailand’s Ministry of Industry, Thailand’s Manufacturing Production Index (MPI) for August 2026 rose 4.44% year on year, with communications equipment growing strongly. However, this came just after the method for calculating the index was revised, so care is needed with simple comparisons against the past.

At the same time, there is cost pressure. The Federation of Thai Industries (FTI) Industrial Sentiment Index fell to 89.7 in August 2026, down from 90.0 in July (Thai PBS World, 9 September 2026). The factors cited for the decline include disruption from heavy rain, weak domestic purchasing power and a shrinking automotive sector, as well as “high energy costs”. On electricity tariffs, Thailand’s Energy Regulatory Commission (ERC) set the fuel adjustment charge (Ft) for September to December 2026 at 16.23 satang per kWh, and the average tariff was held at THB 3.95 per kWh (excluding VAT) (The Nation Thailand, 23 July 2026).

With limited manpower and energy costs on their minds, factories want to know “which machine stops, when, and for how long”, including for older equipment. As a first step, current sensors, which can be fitted without cutting any wiring, come up as a candidate. Retrofit monitoring with current sensors is not a new method. For example, LAPIS Semiconductor announced in April 2018 a relay board for CT sensors to monitor the operating status of existing manufacturing equipment. Precisely because the method has been around for a long time, the typical failure pattern of “installed but never used” is also well known.

Current Sensor Basics: CT, Rogowski and Hall, and Retrofitting with Split-Core Types

The 3 methods, and the difference between AC and DC

An explainer by the measuring instrument manufacturer HIOKI divides current sensors broadly into 3 types.

For retrofit machine monitoring, the type commonly used is the “split-core” CT, whose ring can be partly opened to clamp around a wire. The point to note here is that CTs and Rogowski coils are for AC only. If you want to monitor equipment that runs on DC, or DC circuits, you need a Hall element type or similar. For example, Accuenergy offers split-core Hall-type DC current sensors as products (manufacturer’s nominal values: up to 1,000 A DC, 1% accuracy). Please take this only as an example showing that retrofitting with a split-core type is possible even for DC loads.

A 2024 paper by Madden et al. at Ulster University qualitatively compares shunt resistors, Hall effect sensors, CTs, split-core CTs and Rogowski coils, and adopts the split-core CT as “low cost, with acceptable accuracy, and not requiring machine shutdown for installation”. It cites not needing a shutdown as an advantage over in-line power meters, which require the power to be cut and the wiring disconnected. However, this comparison is qualitative (“high, medium, low”), not a numerical accuracy comparison.

A split-core CT “can be fitted without disconnecting wiring”, but that does not mean “you may work without isolating the power”

The installation guide for Accuenergy’s AcuCT A Series describes the product as intended for “locations where it must be installed around existing conductors or busbars without disconnecting the primary cable”. Depending on the model, the secondary output is either a current output of 80 mA or 100 mA, or a voltage output of 250 mV or 333 mV.

On the other hand, the same guide states that installation and maintenance should be carried out only by qualified personnel with training and experience in high-voltage, high-current equipment, and the first step of its procedure says “where possible, de-energise the circuit, apply lockout/tagout and verify the absence of voltage”. Being able to fit a CT without disconnecting the wiring and allowing anyone to put their hands into a live panel are completely different matters. This point is covered in more detail in the later sections on installation safety and standards and on Thai regulations.

Current Only Tells You “Electricity Is Flowing”: The State Model Is the Real Deliverable

4 ways to read a current waveform

How should CT clamp data be read to arrive at “running”? An explainer by Guidewheel, which provides a current-based machine monitoring service (updated 10 June 2026), says it is read in the following 4 patterns.

• High, stable current: running under load
• Low baseline current: powered on but idle
• Near zero: stopped (powered off)
• Repeating spikes: individual production cycles

This classification is useful, but there is a pitfall here. What current tells you is strictly “how much electricity is flowing”. It does not tell you “why the machine has stopped”. A machine waiting for material, a machine waiting for changeover and a machine whose operator is on a break would likely all look the same, as “idle”, as long as the power is on. If you want to know the reason for a stop, you need a mechanism, separate from current, for entering the reason on the operator panel, a tablet or similar.

Decide the state model first

That is why you should decide in writing “what you call running” before fitting any CT. A study by the Gifu Prefectural Research Institute of Information Technology in Japan (research report for fiscal 2017) defines the states of a surface grinder as the following 3.

This definition changes depending on the factory’s purpose. If you define “powered on = running”, the numbers look good but become hard to use for improvement. If you define “only machining = running”, idle time comes to the surface, and you can act on changeover and material supply problems. Which definition to measure by is decided together with your company’s definition of the utilisation rate (whether planned stops are included, how changeovers are treated). The state model is a “specification” that you will use for longer than the sensor.

Which circuit to fit the CT on

Once the state model is decided, so is where to fit the CT. In the Gifu study, the CT was fitted not on the machine’s main power supply but on the cable on the primary side of the spindle inverter. The study used the property that when the motor is stopped, the only current flowing is what the inverter itself consumes, to determine a stop. Put the other way round, in this configuration the current does not fall to zero even when stopped, as long as the power is on.

Guidewheel says that accuracy problems arise in exceptional cases such as “machines with motors that run constantly regardless of production status”. For example, on equipment whose hydraulic pump is always running, we think the difference between machining and idle may be small if you look only at the main power supply current. In that case, consider fitting the CT not on the main power supply but on the circuit of the spindle or of a load directly linked to machining. Looking at the circuit diagram for each machine and deciding “which wire to clamp so that the difference between the defined states shows up in the current” is the first big hurdle of an implementation.

Whether to measure only 1 phase or all 3 phases of a three-phase machine is also decided here. On machines with unbalanced three-phase loads, it is likely that the value of 1 phase cannot always represent the whole. As a retrofit option other than current, there is also the method of taking signals from indicator lights. A comparison is explained in “Signal Tower Data Collection 2026: A Practical Design for Legacy Machines“.

CT Sensor Threshold Setting: Lessons from a Research Case on Measuring, Smoothing and Timing

A threshold is the rule that converts a current value into a state. Here we look in detail at the Gifu Prefectural Research Institute of Information Technology study of a surface grinder, as a research case of a single machine. The aim is to learn from how the values were decided and verified, not from the values themselves.

Judging on raw current makes the state chatter

In the study, a clamp-type current sensor that can be retrofitted without modification was fitted on the primary side of the spindle inverter; the sensor calculated the RMS value and sent it wirelessly to a gateway. However, because the current had an oscillating waveform, judging it directly against thresholds caused frequent state changes. The study therefore made its judgement using the envelope of the current (the upper and lower outline of the waveform).

This is a problem that can happen in any factory. If raw current is compared directly with a threshold, a momentary dip in current during machining drops the state to “idle”, and it soon returns to “machining”, recording short state changes. Even if the utilisation figure does not change much, the number of stops and minor stops ends up far from reality. As countermeasures, the judgement logic includes measures such as smoothing the waveform with an envelope or moving average, waiting a fixed time before confirming a state, and using different thresholds for rising and falling (hysteresis). Which measures to use, and to what degree, is decided by looking at each machine’s waveform.

The thresholds are values “decided by observing the data”

The study’s judgement rule was: “if the lower envelope is below threshold 1, stopped; otherwise, if the increase in current is above threshold 2, machining; otherwise, idle”. The parameters were 1.0 A for threshold 1, 0.5 A for threshold 2 and a 60-second judgement window, and the report states that they were “determined by observing the time-series data”.

What we want to stress here is that these 1.0 A, 0.5 A and 60 seconds are values decided by looking at real data from a single surface grinder. They are not general values that can be used as they are on other machines. A study by Lv et al. (2016), which measured the power consumption of 4 CNC lathes, 2 milling machines and 1 machining centre, also concluded that the power consumption of non-cutting operations such as idling, and of milling, differs from machine to machine. Thresholds are decided by measuring the baseline for each machine. Rules of thumb such as “stopped if below X% of rated current” sometimes circulate on the internet, but rather than using general values of uncertain origin, it is more reliable to record the current on your own machines for 1–2 weeks and decide by checking it against video or daily reports.

Measure the judgement delay and missed detections

In the study, the start of machining confirmed on video was 16:39:40, while the estimate from current was 16:41:30, a delay of 110 seconds. The study’s discussion attributes this to the grinding wheel being only partly in contact with the workpiece, so the current was small.

The lesson here is that light-load machining can be detected late or missed. A longer judgement window reduces chattering but increases the delay. A shorter one reduces the delay but increases chattering. This trade-off cannot be avoided, so decide first “how much delay is acceptable”, and confirm it by measuring the actual delay in acceptance testing. The deliverable of threshold design includes not just the threshold numbers but also the verification record of “how correctly that threshold judges, and how late”.

Decide the threshold change procedure in advance

Machines change with use. Wear of tools and grinding wheels, changes of product, motor replacement and changes of setup all move the current baseline. If you do not decide who changes thresholds, by what procedure and with what record, the utilisation figure will one day change for no apparent reason. Being able to keep a history of threshold changes is one of the requirements for the system. The approach of making stop records reproducible later is covered in detail in “Equipment Runtime Log Capture: Replayable Stop Analysis“.

Metrics Current Data Can and Cannot Give: OEE Availability, Performance and Quality

Overall equipment effectiveness (OEE) is expressed as Availability × Performance × Quality. In the definition by Vorne Industries (OEE.com), Availability = Run Time ÷ Planned Production Time, Performance = (Ideal Cycle Time × Total Count) ÷ Run Time, and Quality = Good Count ÷ Total Count.

Looking at these formulas makes it clear how far current data can go. If current can determine the state, run time is known, so availability can be calculated. However, performance needs the “total count” and the “ideal cycle time”, and quality needs the “good count”. Guidewheel says production cycles can sometimes be read from repeating spikes, but, as in the Gifu study, light-load machining can be detected late. On quality, Guidewheel also says it is common to supplement with manually entered defect records or integration with existing quality systems.

This article organises it as follows. What CT current alone reliably gives you is mainly “state” and “run time”. Counts and cycles can sometimes be estimated from the waveform, but the good count comes from a separate system.

Let us check with a numerical example (the following are placeholder values for this article). Assume the planned production time of one machine for one day (2 shifts) is 960 minutes.

As a cross-check, OEE can also be calculated as (good count × ideal cycle time) ÷ planned production time, and 700 × 1.0 ÷ 960 = about 72.9%, which matches. If someone proposes “calculating OEE” with CT-only monitoring, always confirm where the performance and quality figures will come from. A configuration that also captures counts and stop reasons corresponds to Configuration B in the model estimate later.

Machine Energy Consumption Measurement: kWh from Current Alone Is a Converted Value

A conversion from current × set voltage, or measured power?

There are frequent requests to measure each machine’s energy consumption (kWh) at the same time as monitoring its operation. What to watch here is what the kWh from a sensor that measures only current actually means. According to the datasheet for Monnit’s ALTA Wireless AC Current Meter, the sensor reports minimum, maximum and average RMS current and ampere-hours (Ah), and can generate Wh and kWh when a default voltage is set on the cloud side. In other words, this kWh is not a value based on measuring the actual voltage, but a conversion from the set voltage. Voltage fluctuations and power factor are likely not reflected, so if you use it for estimating energy savings or allocating costs between departments, use it with that premise understood.

If you want accurate energy figures, you need a power meter that also measures voltage. In that case, the orientation of the CT also becomes important. The Accuenergy installation guide asks that the H1 (or P1) side of the CT faces the source and that all phases are oriented the same way, in order to obtain positive active power, the correct sign of power factor and the correct phase relationship between current and voltage. It states that if negative power or a reversed power factor appears, the CT orientation and the voltage phase assignment should be checked. For operation judgement that looks only at the magnitude of current, the effect of orientation is likely small, but it is essential if you measure power. How to collect measured kWh at the incoming supply and main panels is explained in “Power Meter Data Collection: An RFP for Trusted 15-Minute Data“.

Choosing the measuring range and rating

A CT has a range it can measure. For example, the measuring range of the AcuCT A Series is said to be 5% to 120% of rated current (manufacturer’s nominal values). If you choose a CT with too large a rating, the small current when idle may fall below the bottom of the range, making it hard to tell stopped from idle. Choose the CT rating by looking not at the capacity of the main circuit breaker, but at the actual current in the states you want to measure (especially the idle current).

Reporting interval and battery life of wireless battery-powered types

To reduce wiring work, you may choose battery-powered wireless CT sensors. Here there is a trade-off between reporting interval and battery life. Monnit’s datasheet gives the battery life as “10+ years on AA batteries”, but notes that it depends on conditions such as reporting frequency (manufacturer’s nominal values). The internal memory holds up to 512 readings, and how many days that covers also depends on the reporting interval (in the same datasheet, 3.5 days at a 10-minute interval and 42 days at a 2-hour interval). The paper by Madden et al. points out, for a different battery-powered wireless node, that data granularity is limited by battery life, and that a 10-minute reporting interval may be too coarse to capture fine changes in current in production processes.

The time resolution needed for operation judgement is determined by the “acceptable delay” decided in threshold design and by how short the stops you want to count are. If you want to count minor stops, a mains-powered configuration that can send fine-grained data may suit better than a battery-powered one. As a product example, HARDWARIO’s CHESTER Current is described as having up to 4 channels of split-core clamp-type probes, measuring AC and DC current and transmitting over NB-IoT, LTE-M and LoRaWAN (manufacturer’s nominal values; IP67, operating temperature −20 to +60°C). If you use wireless products, please confirm that they have obtained radio certification in the country where they will be deployed.

Installation Safety and Standards: Qualified Personnel, Isolation and Lockout, Open Secondary, Single-Conductor Clamping

Panel work by qualified personnel, with isolation and lockout

As mentioned earlier, the installation guide for the AcuCT A Series limits installation and maintenance to qualified personnel, and asks that work be done with the power isolated and lockout/tagout applied wherever possible, and that the absence of voltage be verified. Where live work is necessary, it states that the live working procedure approved at that facility should be followed, using arc flash protective equipment and insulated tools. Having maintenance staff open a live power panel and work inside “because it is just clamping something on” should be avoided. Regulations in Thailand are organised later in “Issues Specific to Thailand and ASEAN”.

Do not open-circuit the secondary of a current-output CT

Some CTs require care in how their secondary side is handled. Several technical explainers state that, in general, if the secondary of a current-output CT is opened while current is flowing in the primary, a dangerously high voltage is induced at the secondary terminals. On the other hand, the AcuCT A Series guide says that all models have internal open-circuit protection and the secondary does not produce a dangerous open-circuit voltage, so shorting blocks are not required, and that the open-circuit risk of voltage-output products and the like is lower than that of current-output CTs (though it says the wiring should still be terminated and protected). This statement applies only to that product. Check whether the CT you use is current-output or voltage-output and whether it has built-in protection, and for current-output CTs without protection, write a procedure into the work instructions not to open-circuit the secondary.

Pass only 1 conductor of 1 phase through it

A surprisingly common failure is in how the CT is clamped. Monnit’s datasheet asks that the CT be fitted around only 1 wire of an AC circuit, and says that if the outgoing and return wires are clamped together, the current reads 0. The Accuenergy guide also says the CT should be fitted around only 1 conductor of 1 phase, and that conductors of multiple phases should not pass through the same CT. Outgoing and return currents cancel each other out, so clamping a cable with several cores bundled together will not give a correct value. In the panel, choose wires that are separated by phase and clamp those. The same guide asks that secondary wiring be kept away from sharp edges, moving parts, hot parts and noisy conductors, and kept as short as possible.

Standards and measurement categories

Let us also sort out how standards apply. For instrument current transformers, there is IEC 61869-2 (additional requirements for current transformers, 2012), which is said to have replaced IEC 60044-1 and others. On the other hand, IEC 61010-2-032 covers hand-held current clamps, and the IEC Webstore description of an earlier edition excludes fixed-installation current transformers and current transducers. In other words, CTs permanently installed in panels for machine monitoring and the clamp meters that maintenance staff carry around belong to different worlds of standards. Low-cost CTs for machine monitoring do not necessarily all conform to IEC 61869-2, so in the RFP, ask suppliers to show the applicable standards in the manufacturer’s documentation.

Also check the measurement category. In Fluke’s explainer (based on IEC 61010-1), CAT II covers single-phase loads connected to outlets, CAT III covers three-phase distribution (including fixed-installation equipment such as switchgear and polyphase motors), and CAT IV covers the origin of installation, outdoor conductors, utility meters and service entrances. A factory’s power panels and machine main supplies are at positions equivalent to CAT III, so choose products of a category that suits the installation position (for example, the AcuCT A Series is said to be 600 VAC, CAT IV; manufacturer’s nominal values).

Cost and ROI: Estimating a Current Sensor Machine Monitoring Implementation at a Model Factory

All figures from here on, except the average electricity tariff, are placeholder values set independently by this article. They are neither industry averages nor survey results. Because there is no publicly available primary information on the prices of sensors, gateways, installation work and integration, the amounts are placeholder values. For actual decisions, please recalculate using quotations from several companies and your own track record.

Common assumptions (Model Factory M)

Configuration A: State monitoring with CTs only (20 machines)

Initial investment is CT sensors and transmitter nodes for 20 machines × THB 25,000 = THB 500,000; gateway and server (or cloud) setup THB 150,000; panel installation work (by qualified personnel, including power isolation, lockout and updating circuit diagrams) 20 machines × THB 8,000 = THB 160,000; and threshold design for each machine (baseline measurement and state model definition) plus SAT, THB 190,000. The total is 500,000 + 150,000 + 160,000 + 190,000 = THB 1,000,000. Annual operating cost for cloud use and maintenance is set at THB 120,000 per year.

Benefits (assumed values) are as follows.

• Holiday work: 24 → 14 days. Because stop times become visible, action can be taken on changeovers and material waits. The reduction of 10 days × THB 30,000 = THB 300,000
• Idle power: identify, from state data, the time machines sit “powered on and idle” during lunch breaks, between shifts and after closing, set rules for switching off, and cut 2 hours per machine per day. 20 machines × 2 hours × 300 days = 12,000 hours, × 2 kW = 24,000 kWh, × THB 3.95 = THB 94,800
• Staff: production counts remain handwritten, so the daily report staff reduction is 0

Total benefits are 300,000 + 94,800 = THB 394,800 per year, annual net benefit is 394,800 − 120,000 = THB 274,800 per year, and the simple payback period is 1,000,000 ÷ 274,800 = about 3.6 years.

Configuration B: CT + PLC and counter integration to capture counts automatically (20 machines) + stop reason entry

Initial investment adds the following to the full Configuration A package of THB 1,000,000 (including threshold design and SAT): PLC integration (8 machines) 8 × THB 40,000 = THB 320,000; adding counter signals to the 12 old machines (cycle-complete signals, proximity sensors, etc.) 12 × THB 15,000 = THB 180,000; 4 tablets for entering stop reasons × THB 25,000 = THB 100,000; daily report automation and integration with existing systems THB 200,000; and FAT/SAT for the additions THB 200,000. The additions come to 320,000 + 180,000 + 100,000 + 200,000 + 200,000 = THB 1,000,000, and the total is THB 2,000,000. Annual operating cost is set at THB 200,000 per year.

Benefits (assumed values) are calculated as replacing the benefits of Configuration A (they are not added to Configuration A’s benefits).

• Holiday work: 24 → 10 days. Because even the reasons for stops are known, countermeasures are faster. The reduction of 14 days × THB 30,000 = THB 420,000 (including Configuration A’s 10 days)
• Idle power: THB 94,800, the same as Configuration A (the state data comes from the same CTs, so it does not increase in B)
• Staff: compiling and transcribing daily reports is no longer needed. 1 person × THB 240,000 = THB 240,000

Total benefits are 420,000 + 94,800 + 240,000 = THB 754,800 per year, annual net benefit is 754,800 − 200,000 = THB 554,800 per year, and the simple payback period is 2,000,000 ÷ 554,800 = about 3.6 years.

Key point 1: Payback years are almost the same. B comes out ahead on the 5-year cumulative figure

The simple payback period is about 3.6 years for both, almost the same. The 5-year cumulative figures are A: 274,800 × 5 − 1,000,000 = 374,000 and B: 554,800 × 5 − 2,000,000 = 774,000, so B comes out ahead.

Taking only Configuration B’s increment, the additional initial investment is 2,000,000 − 1,000,000 = THB 1,000,000, the additional annual net benefit is 554,800 − 274,800 = THB 280,000 per year, and the payback on the increment is 1,000,000 ÷ 280,000 = about 3.6 years. The difference in the 5-year cumulative figures is 280,000 × 5 − 1,000,000 = THB 400,000, which matches 774,000 − 374,000.

This increment is the value of “counts and stop reasons”. It corresponds to the point made in the earlier section that current alone cannot give performance (counts) and cannot tell you the reasons for stops. Which one to start with is decided by the effort spent compiling daily reports and by your record of holiday work. For a factory where the daily report workload is small, it is also a rational sequence to start with Configuration A to make states visible, confirm the benefits, and then move on to capturing counts.

Key point 2: What if threshold design and SAT were left out?

What happens if, because CTs are cheap, you cut the THB 190,000 for threshold design and SAT? Let us call the configuration with this cost removed from Configuration A “Configuration A′”.

Initial investment falls to 1,000,000 − 190,000 = THB 810,000. However, we assume that data whose state judgement chatters and which cannot reliably tell idle from machining will not be trusted on the shop floor and will stop being used for countermeasures. Holiday work only falls from 24 → 22 days, giving 2 days × THB 30,000 = THB 60,000. If the switch-off rules for idle power are also only half applied, that is 94,800 ÷ 2 = THB 47,400. Total benefits are 60,000 + 47,400 = THB 107,400 per year, and after subtracting the operating cost of THB 120,000, the result is −THB 12,600 per year. It does not even cover the operating cost, and the investment never pays back.

Cut THB 190,000 and the annual net benefit falls from THB 274,800 to −THB 12,600. This is the single most important message of this estimate. Of course, how far the benefits fall is a placeholder assumption, but the structure that “CTs are cheap, but what decides success is threshold design and acceptance testing” is likely the same in any factory.

Cross-check notes (avoiding double counting)

Configuration B’s benefits replace Configuration A’s benefits, and the two must not be added together. The reduction of 14 holiday work days in B includes A’s 10 days, and the THB 94,800 of idle power is the same in A and B. There is only 1 starting point: the current situation of 24 holiday work days per year, THB 30,000 per day, 1 daily report staff member and 2 kW of idle power. Also, kWh obtained with CTs only is a conversion from current and a set voltage, so the idle power benefit is treated as a rough estimate at the planning stage.

12 Items to Write in a Machine Monitoring System RFP

These are the items you should write, at a minimum, in an RFP (request for proposal) when asking manufacturers and system integrators for quotations.

1. Equipment register and load types: model of each machine, power supply (three-phase or single-phase, AC or DC), main loads (spindle, hydraulic pump, heaters, etc.), whether an inverter is fitted
2. State model definition: definitions of states such as stopped, idle and machining, and the company’s definition of the utilisation rate (treatment of planned stops and changeovers)
3. CT mounting circuit, rating and method: main power supply or spindle/load side, 1 phase or 3 phases, a rating suited to the actual idle current, CT for AC and Hall type or similar for DC
4. How thresholds are decided, and the change procedure and authority: baseline measurement period, judgement logic (envelope, moving average, hysteresis, judgement window), recording the history of threshold changes
5. Sampling and reporting intervals, and judgement delay: acceptable delay, the shortest stop you want to count, retention in internal memory during communication outages
6. kWh calculation method: measured power (including voltage and power factor), or a conversion from current × set voltage. If power is measured, CT orientation and voltage phase assignment
7. How counts and stop reasons are captured: whether by PLC integration, counter signals or input terminals. Whether integration with quality records is needed
8. Wireless method, radio certification and batteries: communication method, radio certification in the country of deployment, and for battery types, battery life against reporting interval (manufacturer’s nominal values and conditions)
9. Data storage, export and integration with existing systems: retention period, export such as CSV, connection to production management systems and MES, time synchronisation
10. Arrangements for panel work: work by qualified personnel, power isolation planning with lockout and signage, updating electrical circuit diagrams after the work, inspection and maintenance records
11. Local support and Thai-language support: maintenance contact within Thailand, replacement when a sensor fails, explanations in Thai for shop-floor staff
12. FAT/SAT criteria: test items, how reference data is captured (video, records), acceptance criteria, scope of witnessing

Items 2, 4 and 6 are especially important. If 2 remains vague, you end up comparing proposals in which each company uses “running” to mean something different. If you do not write 4, you get a system where thresholds are set once and nobody can fix them. If you do not check 6, converted kWh gets treated as if it were measured. How to position machine monitoring within company-wide manufacturing data collection is organised in “Manufacturing Data Collection: Thailand 90-Day PoC and RFP“.

What to Check at Machine Monitoring FAT/SAT

In the factory acceptance test (FAT) before shipment and the site acceptance test (SAT) after installation, confirm not the catalogue values but that, on your own machines, the system “judges correctly, keeps the delay within the acceptable range and does not lose data”. Current-based judgement gives different results for each machine and circuit, so SAT is particularly important.

• Rating and actual current range: check that the actual currents when idle and when machining fall within the CT’s measuring range (manufacturer’s nominal values)
• State judgement accuracy: using video or work records over a set period as the reference, compare them with the system’s judgements and record the proportion of time that matched for each state
• Delay when the state changes: at the start and end of machining, measure the difference between the time on video and the time of judgement, and see whether it falls within the tolerance set in the RFP
• Chattering: count how many short state changes occur and check that the number of stops does not diverge from reality
• Conditions prone to misjudgement: deliberately create conditions such as light-load machining, machines with constantly running motors and dry running during changeovers, and check the judgement
• Clamping, orientation and phase: check that each CT is on 1 conductor of 1 phase, and, if power is measured, that CT orientation and voltage phase assignment are correct (that no negative power appears)
• Communication outages and data loss: temporarily stop communication and check retention in internal memory, and resending and the display of gaps after recovery
• Recovery and time after a power outage: check that measurement resumes automatically after a power cut and re-energisation, and that the clock has not drifted

Measuring state judgement accuracy takes effort, but if you skip it, as seen in Key point 2, the data will not be trusted and the investment will not pay back. How to capture reference data has a lot in common with verifying short stops, so please also see “Minor Stop Reduction: 30-Day Monitoring PoC in Thailand“.

Issues Specific to Thailand and ASEAN

Thailand’s ministerial regulation on electrical occupational safety and health (B.E. 2558)

Thailand’s Ministry of Labour has a ministerial regulation setting standards for occupational safety and health relating to electricity (Buddhist Era 2558, published in the Royal Gazette on 6 February 2015). We introduce the key points of the provisions as summarised by this article. The regulation defines workers who install, inspect, test, repair and maintain electrical systems, electrical equipment and wiring as “workers performing electrical work”. It then places obligations on employers such as the following.

• Have workers performing electrical work receive training (Section 4)
• Keep the electrical circuit diagram of the workplace, certified by an engineer licensed in the electrical field or by the regional electricity authority, and correct the diagram properly if there is any modification or change (Section 5)
• Records of inspection and maintenance of electrical equipment are prepared and certified by a person registered under the law or a licensed juristic person (Section 12)
• During work on electrical systems or equipment, take preventive measures such as locking so that switches cannot be turned on, and display a sign prohibiting switching on (Section 15)

Work to add CTs inside a panel may count as “installation”, and Section 15’s lockout and signage and Section 5’s updating of circuit diagrams could be relevant. However, the legal judgement of which work falls under which provision should be confirmed individually with the factory’s safety officer, electrical engineers and specialists. Including the updating of circuit diagrams after the work in the RFP item “Arrangements for panel work” will save you a scramble later.

Skill certification for electricians

According to a Thai PBS article from October 2019, Thailand’s Department of Skill Development (DSD), under the Skill Development Promotion Act, designates indoor electricians as an “occupation in which only persons who have received a skill certificate may work”. However, this article concerns indoor electricians, and we have not been able to confirm whether industrial electrical work in factories is covered. As it is also a 2019 article, check the latest designated occupations with the DSD, and ask companies to which you subcontract the work to show the qualifications of their workers.

Electricity tariffs differ by contract

The estimate used the Thailand average tariff of THB 3.95/kWh (excluding VAT) for September to December 2026 indicated by the ERC. Actual factory tariffs differ by contract type, such as TOU and large general service. Note that, according to The Nation (12 August 2026), the tariff for September billing was cut to an average of THB 3.89/kWh, but this includes measures aimed at households, and this article has not been able to confirm how it applies to industrial customers. When calculating energy savings, please use the tariff on your own electricity bill. If you suspect machines are stopping because of power quality problems such as voltage sags, look at voltage as well as current. The approach is explained in “Voltage Monitoring System for Thailand Factory Power Quality“.

Heat, humidity and panel temperature, and radio certification

The inside of panels in Thai factories can get hot. Check the operating temperature range of sensors and transmitter nodes in the manufacturer’s nominal values (for example, the AcuCT A Series is said to be −25 to 70°C, and CHESTER Current −20 to +60°C). Even if wireless products have overseas certifications, they do not necessarily have certification for use in Thailand. Please confirm with the manufacturer or distributor whether they have radio certification in the country of deployment.

The BOI’s upgrading measure

The BOI’s investment promotion guide (2023 edition), under the “Measure for Industrial Upgrades towards Smart and Sustainable Industry”, sets out, for efficiency improvements through machinery upgrades and automation in existing businesses, conditional on an investment of at least THB 1 million (excluding land and working capital), exemption of import duty on machinery and a 3-year corporate income tax exemption (capped at 50% of the upgrade investment, or 100% if linkage with Thailand’s automation industry exceeds a certain level). Investment in software and IT integrated with machinery to control, monitor and support manufacturing processes is also listed as counting towards the investment amount. However, this is a 2023 document and the conditions may have changed. In addition, this article has not been able to confirm whether retrofitting CT sensors alone would qualify. Please confirm the latest conditions and eligibility individually with the competent authority (the BOI) or specialists.

Examples of options on the market

Examples of current sensor and machine monitoring products and services include Accuenergy for split-core CTs and DC sensors, Monnit for wireless current meters, HARDWARIO (CHESTER Current) for wireless nodes with clamp-type probes, and Guidewheel for machine monitoring services using CT clamps. The measuring instrument manufacturer HIOKI publishes an explainer on current sensor methods. There are many other products as well. These are listed only as examples and are not recommendations or comparative evaluations. If you are considering IoT for old equipment together with decisions to retain, migrate or retire it, “Aging Equipment IoT: Retain, Migrate or Retire” may also be useful.

90-Day Plan

Days 0–30: Inventory of target equipment, compiling the current situation, and defining the state model

List the machines you want to cover and check their model, power supply, main loads, whether an inverter is fitted and whether they can connect to a PLC. Also check at this stage whether the panel circuit diagrams match the current state. At the same time, pull out the number and cost of holiday work days over the past year and the effort spent compiling daily reports, and replace the assumptions of the model estimate with your own values. Finally, document the state model, such as “stopped, idle, machining”, and the company’s definition of the utilisation rate.

Days 31–60: Measure the baseline on representative machines and provisionally set thresholds

For each type of machine, choose 1–2 representative units, have CTs fitted by qualified personnel, and record the current for 1–2 weeks. Over the same period, take video or work records, compare the current waveform with the actual states, and provisionally decide the thresholds and judgement logic (how to smooth the waveform, the judgement window). Also check here how late light-load machining is detected, and whether idle and machining can be separated on machines with constantly running motors. For machines where they cannot be separated, consider changing the mounting circuit, adding a count signal or another method.

Days 61–90: RFP, FAT/SAT criteria and the order decision (A or B)

Based on the measurement results, finalise the 12 RFP items and the FAT/SAT test items and acceptance criteria (accuracy and acceptable delay), and obtain quotations from several companies under the same conditions. Using your own version of the estimate, decide whether to start with Configuration A (CT only) or go on to Configuration B (including counts and stop reasons).

The deliverables you should have in hand at the end of the 90 days are 5: (1) the equipment register and the results of the circuit checks, (2) the state model and the definition of the utilisation rate, (3) thresholds for the representative machines and the judgement verification records (accuracy, delay), (4) your own version of the investment estimate, and (5) the RFP and FAT/SAT criteria.

Frequently Asked Questions

What can and cannot be known when monitoring machines with current sensors (CT clamps)?

What current tells you is how much electricity is flowing. With threshold-based judgement, you can know states such as stopped, idle and machining, and their durations (availability). On the other hand, you cannot know why a machine has stopped (waiting for material, waiting for changeover, etc.), and production counts (performance) and good counts (quality) need separate data. Counts can sometimes be estimated from the waveform, but light-load machining can be detected late. The realistic approach is to supplement stop reasons with input terminals, counts with PLCs or counters, and good counts with quality records.

Can machine monitoring be retrofitted to existing equipment? Is a shutdown needed?

Split-core CTs can be fitted to existing wires without disconnecting the primary wiring, so they can be retrofitted without modifying the machine. However, not disconnecting the wiring is different from allowing anyone to work inside a live panel. The manufacturer’s installation guide asks that qualified personnel do the work, isolating the power and applying lockout/tagout wherever possible, and verifying the absence of voltage. In Thailand, the ministerial regulation on electrical occupational safety and health (B.E. 2558) sets out requirements such as locking and signs prohibiting switching on during work, and updating circuit diagrams. Please confirm the specific application individually with the factory’s safety officer, electrical engineers and specialists.

How should CT sensor thresholds be set?

Decide them by measuring the current baseline for each machine. Power consumption when idle or not cutting differs between machines, so using general rule-of-thumb values as they are is not recommended. In the study by the Gifu Prefectural Research Institute of Information Technology, the thresholds and judgement window for a single surface grinder were decided by observing time-series data, and the waveform was smoothed using an envelope to suppress state chattering. Even so, detection of light-load machining was delayed. Measure accuracy and delay by comparing against video or work records, and also decide the threshold change procedure and history.

Can current sensors also measure energy consumption (kWh)?

Some products can produce kWh by converting from a set voltage, even with sensors that measure only current. However, these are not values based on measuring actual voltage or power factor, so treat them as rough estimates. If you need accurate energy figures, use a power meter that also measures voltage, and align CT orientation (H1/P1 towards the source) and voltage phase assignment. In the RFP, have suppliers state whether kWh is measured power or a conversion.

What should be checked in a machine monitoring system RFP and at FAT/SAT?

In the RFP, write items such as the state model definition, CT mounting circuit and rating, how thresholds are decided and the change procedure, acceptable judgement delay, kWh calculation method, how counts and stop reasons are captured, and arrangements for panel work. At FAT/SAT, check state judgement accuracy using video or work records as the reference, delay when the state changes, the number of chattering events, behaviour under conditions prone to misjudgement, clamping and orientation, and recovery after communication outages and power cuts.

What are the cost and payback period of current sensor machine monitoring?

There is no publicly available primary information on sensor and installation prices, so this article’s model estimate was calculated with placeholder values. Configuration A, monitoring 20 machines with CTs only, had an initial investment of THB 1,000,000 and a simple payback of about 3.6 years; Configuration B, which also captures counts and stop reasons, had THB 2,000,000 and about 3.6 years; and the 5-year cumulative figures were THB 374,000 for A and THB 774,000 for B. The calculation also shows that leaving out threshold design and SAT makes the annual net benefit negative, so the investment never pays back. Please recalculate with your own track record and quotations from several companies. Whether you qualify for BOI promotion measures should be confirmed individually with the competent authority and specialists.

Summary

• What current tells you is that “electricity is flowing”, not that “the machine is producing”. The real deliverables are the state model (the definitions of stopped, idle and machining) and the threshold for each machine.
• CTs are AC-only and should have 1 conductor of 1 phase passed through them. Which circuit you fit them on, main power supply or spindle/load side, changes how easy judgement is. Take particular care with machines that have constantly running motors.
• Decide thresholds by measurement on each machine, suppress chattering and measure the judgement delay. The values in the research case are from a single machine and are not general values.
• What current gives you is mainly availability. Performance needs counts, quality needs good counts, and stop reasons need an input mechanism. kWh from current alone is a converted value.
• Panel work is done by qualified personnel, including power isolation, lockout and updating circuit diagrams. How Thailand’s ministerial regulation and qualification requirements apply should be confirmed individually.
• In the model estimate (placeholder values), both CT-only and count integration pay back in about 3.6 years. Cut threshold design and SAT and the investment never pays back. What decides success is not the price of the sensor but threshold design and acceptance testing.

TOMAS TECH supports Japanese-owned factories in Thailand, from preparatory stages such as taking inventory of target equipment, defining the state model and designing thresholds on representative machines, through to drafting the RFP, arranging panel work, witnessing FAT/SAT and integrating with existing systems. Even if you are at the stage of “first wanting to sort out which circuit on which machine to fit sensors to”, please feel free to get in touch via our contact form.

References

• HIOKI, “Current Sensor Working Principle & Characteristics”: https://www.hioki.com/id-id/learning/test-tools/general-current-sensing.html
• Accuenergy, “DC Current Sensors”: https://accuenergy.com/products/dc-sensors
• Accuenergy, “AcuCT A Series Revenue Grade Split-Core Current Transformer Installation Guide” v1.0.1: https://www.accuenergy.com/wp-content/uploads/acuct-a-series-revenue-grade-split-core-current-transformer-installation-guide.pdf
• Madden, Coleman, Kerr, Quinn, “Retrofitting Manufacturing Efficiency: Low-Cost Electrical Current Monitoring for Industrial Machinery” (IET LoDiSA 2024, 3 October 2024): https://pure.ulster.ac.uk/ws/files/216763877/madden-et-al-2024-retrofitting-manufacturing-efficiency-low-cost-electrical-current-monitoring-for-industrial-machinery.pdf
• Industrial Monitor Direct, “Current Transformer Open Secondary Terminal Hazards”: https://industrialmonitordirect.com/blogs/knowledgebase/current-transformer-open-secondary-terminal-hazards
• Monnit, “ALTA Wireless AC Current Meters” datasheet (distributed by Omega): https://assets.omega.com/spec/AC%20Current%20Meter%20Data%20Sheet.pdf
• Gifu Prefectural Research Institute of Information Technology, Research Report No. 19, “Research and Development on Preventive Maintenance Using IoT Technology (Report 1)”: https://www.gitec.rd.pref.gifu.lg.jp/files/reports/2017/information_2017_01.pdf
• Guidewheel, “Machine monitoring without IT, no PLC” (updated 10 June 2026): https://www.guidewheel.com/blog/machine-monitoring-without-it-no-plc
• Lv J. et al., “Experimental study on energy consumption of computer numerical control machine tools” (Journal of Cleaner Production, 2016): https://eprints.gla.ac.uk/145041
• Vorne Industries (OEE.com), “Calculating OEE”: https://www.oee.com/calculating-oee/
• VDE Verlag, “IEC 61869-2 Ed.1.0” preview: https://www.vde-verlag.de/iec-normen/preview-pdf/info_iec61869-2{ed1.0}b.pdf
• IEC Webstore, “IEC 61010-2-032”: https://webstore.iec.ch/en/publication/18841
• Fluke, “IEC Category Ratings”: https://my.fluke.com/en-us/article/KA-03314
• HARDWARIO, “CHESTER Current”: https://www.hardwario.com/products/chester/catalog/current/
• LAPIS Semiconductor, “CT Sensor Shield 2” press release (27 April 2018): https://digitalpr.jp/r/26982
• Thailand Ministry of Labour, ministerial regulation on electrical occupational safety and health (B.E. 2558, copy of the Royal Gazette): https://ams.medsci.nu.ac.th/wp-content/uploads/2022/07/LAW2022/S13-%20กฎกระทรวง%20ความปลอดภัย%20อาชีวอนามัย%20และสภาพแวดล้อมในการทำงานฯ%202558.pdf
• Enviliance, “Thailand OSH”: https://enviliance.com/regions/southeast-asia/th/th-osh
• Thai PBS (DSD skill assessment for electricians, 21 October 2019): https://www.thaipbs.or.th/news/content/285360
• Thailand BOI OSOS (investment applications in the first half of 2026, 23 July 2026): https://osos.boi.go.th/EN/news/2430/Thailand-Secures-43-6bn-1H-2026-Investment-Surge-as-Big-Tec/
• BOI, “Measure for Industrial Upgrades towards Smart and Sustainable Industry” (excerpt from the 2023 investment promotion guide): https://www.boi.go.th/upload/content/Smart_and_Sustainable_Industry.pdf
• The Nation Thailand, “Power tariff held at 3.95 baht through year-end” (23 July 2026): https://www.nationthailand.com/news/40068945
• The Nation Thailand, “Thailand cuts September power tariff, caps first 200 units at 3 baht” (12 August 2026): https://www.nationthailand.com/news/40069705
• Trading Economics, “Thailand Industrial Output Growth Beats Forecasts” (30 September 2026): https://tradingeconomics.com/thailand/industrial-production/news/587981
• Thai PBS World (FTI Industrial Sentiment Index, 9 September 2026): https://www.thaipbsworld.com/world/557682