Thermoelectric Energy Harvester and Piezoelectric Vibration Monitor Project

This project designs a self-powered vibration monitoring node for a warm rotating machine. A thermoelectric generator harvests heat from the machine casing, a storage element buffers the measurement and radio bursts, and a piezoelectric accelerometer measures vibration through a charge amplifier and sampled signal chain.

The engineering goal is not to prove that thermoelectric and piezoelectric effects exist. The goal is to produce a design review package showing that the available thermal gradient can power the node, the piezoelectric channel can measure the required vibration without clipping or aliasing, and the installed system can be validated before release.

Project Objective

Produce a design and validation package for a condition-monitoring node that answers:

1. Is enough thermoelectric power available after thermal interface losses?
2. Can the storage element support startup, measurement bursts and radio transmission?
3. Does the piezoelectric measurement chain have enough range, resolution and bandwidth?
4. Does the mounted sensor resonance stay outside the diagnostic band?
5. Which tests prove that the installed system meets the requirements?
6. Which evidence should be included in the release report?

The final deliverable is a compact engineering decision package: requirements, architecture, calculations, validation matrix, acceptance criteria, risks and release recommendation.

System Requirements

Use these project requirements as the baseline.

Requirement	Target	Engineering reason
Machine casing temperature	65\ ^\circ\text{C} nominal	Defines hot-side opportunity for the thermoelectric generator.
Ambient temperature	25\ ^\circ\text{C} nominal	Defines external thermal gradient.
Required diagnostic band	10 Hz to 5 kHz	Covers imbalance, bearing and general casing-vibration indicators.
Expected normal peak acceleration	8g	Sets normal piezoelectric signal level.
Credible startup shock	30g	Sets overload margin.
Sampling rate	20 kS/s	Keeps the 5 kHz band below Nyquist with room for filtering.
Measurement burst	2 s every 60 s	Defines sensor and processor energy duty cycle.
Radio burst	0.3 s every 60 s	Defines communication energy duty cycle.
Continuous sleep load	3 mW	Defines baseline energy drain.
Release evidence	thermal, vibration, power and firmware tests	Prevents a calculation-only release.

These numbers should be adapted to the actual machine, radio protocol, enclosure, mounting method and diagnostic objective.

Proposed Architecture

The node contains:

1. thermoelectric generator clamped between machine casing and heat sink;
2. thermal interface material and controlled clamp load;
3. boost or buck-boost energy harvester with maximum power point behavior;
4. storage capacitor or rechargeable storage element;
5. piezoelectric accelerometer on the bearing housing or casing location of interest;
6. shielded cable or integrated low-noise connection;
7. charge amplifier with selectable feedback capacitance and feedback resistance;
8. anti-alias filter and ADC;
9. microcontroller performing burst acquisition and feature extraction;
10. radio transmitter;
11. firmware watchdog, brownout handling and data-quality flags;
12. validation record for installation, calibration and release.

The architecture must be reviewed as one coupled system. A thermal interface change can reduce available power. A charge-amplifier gain change can prevent clipping but degrade resolution. A firmware duty-cycle change can invalidate the storage sizing.

Input Data for the Design Review

Thermoelectric and power data:

Quantity	Symbol	Value
external temperature difference	\Delta T_{available}	40\ \text{K}
heat flow through thermoelectric stack	Q	8.0\ \text{W}
hot-side interface resistance	R_{hot}	1.0\ \text{K/W}
cold-side interface resistance	R_{cold}	1.2\ \text{K/W}
module Seebeck coefficient	S	0.032\ \text{V/K}
module internal resistance	R_i	4.0\ \Omega
harvester conversion efficiency at operating point	\eta_h	65 percent

Node load data:

Load item	Power	Duration and duty
sleep electronics	3\ \text{mW}	continuous
measurement and processing burst	250\ \text{mW}	2 s every 60 s
radio burst	800\ \text{mW}	0.3 s every 60 s
cold-start check burst	300\ \text{mW}	5 s

Piezoelectric channel data:

Quantity	Symbol	Value
accelerometer charge sensitivity	S_q	250\ \text{pC}/g
selected feedback capacitance	C_f	4.7\ \text{nF}
amplifier linear output limit	V_{lim}	\pm 4.5\ \text{V}
ADC input span	FS	\pm 5\ \text{V}
ADC resolution	N	16 bit
first mounted resonance target	f_0	35\ \text{kHz}
maximum diagnostic frequency	f_{max}	5\ \text{kHz}

Step 1: Derate the Thermoelectric Temperature Difference

The external machine-to-ambient temperature difference is not the same as the temperature difference across the thermoelectric module. Interface losses consume part of it.

Interface temperature loss:

Module temperature difference:

Engineering Comment

Only 22.4\ \text{K} remains across the module. A design based on the external 40\ \text{K} would overstate power. The project should specify interface material, surface finish, clamp force and heat-sink orientation because these details control the usable thermal gradient.

Step 2: Estimate Thermoelectric Power

Open-circuit voltage:

For a matched electrical load:

So the ideal matched module power is:

Usable harvested power after conversion:

Engineering Comment

The harvester is plausible only if the node average power stays below about 20.9\ \text{mW} with margin. The open-circuit voltage is also low, so the selected power-management circuit must be able to cold-start and operate at this voltage range.

Step 3: Check Average Power Budget

Continuous sleep load:

Measurement burst average:

Radio burst average:

Total average node power:

Power margin:

Engineering Comment

The average power margin is only 36 percent. That is acceptable for a controlled prototype but thin for production because ambient temperature, machine load, dust on the heat sink, interface aging and radio retries can reduce margin. A release design should include lower-power firmware modes or a larger thermal harvester if the machine will not always run at the nominal temperature.

Step 4: Size Storage for a Cold-Start Burst

The cold-start check requires:

for:

Required energy:

A storage capacitor is evaluated between:

and:

Available capacitor energy is:

For:

the stored usable energy is:

Engineering Comment

The storage element can support the 1.50\ \text{J} cold-start burst with a large energy margin. This does not prove startup reliability. The harvester must still charge the capacitor from the worst-case initial state, the converter must operate at the low voltage, and firmware must avoid repeated failed start attempts that drain storage.

Step 5: Check Piezoelectric Charge-Amplifier Range

Charge-amplifier voltage sensitivity is:

With:

and:

the result is:

Output at normal peak acceleration:

Output at credible startup shock:

Approximate clipping acceleration:

Engineering Comment

The selected feedback capacitance gives enough overload margin for the specified 30g startup shock. It also leaves normal vibration far below the amplifier limit. If the diagnostic requirement later demands very low vibration resolution, the team may need a higher-gain range or a second channel.

Step 6: Check ADC Resolution

The ADC span is:

for a \pm 5\ \text{V} input range. With 16 bits:

Acceleration code width:

Engineering Comment

The quantization step is about 0.0029g before noise, filtering and calibration uncertainty. That is adequate for many casing-vibration alarms, but final resolution must be checked using measured noise with the sensor mounted and the machine off, then with the machine running.

Step 7: Check Sampling and Mounted Resonance

The sampling rate is:

Nyquist frequency:

The diagnostic band reaches:

so the measurement band is below Nyquist. The analog anti-alias filter should attenuate strongly before 10\ \text{kHz} while preserving the diagnostic band.

For a simple mounted-resonance margin:

The maximum diagnostic frequency is:

Engineering Comment

The resonance margin passes this screening rule. The project should still verify the mounted resonance because adhesive thickness, mounting torque, local casing stiffness and sensor mass can shift it. A shaker or instrumented comparison test should confirm the usable band after installation.

Validation Matrix

Validation item	Method	Acceptance evidence
Thermal interface	Thermal soak at low, nominal and high machine temperatures	Measured hot-side, cold-side and ambient temperatures; calculated module \Delta T.
Harvested power	Run with representative duty cycle	Storage voltage remains above brownout threshold over a full operating cycle.
Cold start	Start from discharged storage under minimum thermal gradient	Node boots, completes health check and records first valid data packet.
Piezo range	Shaker or controlled vibration input plus startup shock simulation	No clipping at expected shock; normal band has usable resolution.
Sampling chain	Sine sweep and anti-alias check	Gain and phase response documented over 10 Hz to 5 kHz.
Mounted resonance	Installed modal or sweep test	First mounted resonance remains outside the release limit.
Firmware duty cycle	Power logging and brownout test	No repeated restart loop; watchdog and data-quality flags work.
Environmental robustness	Temperature cycling, cable movement and heat-sink contamination check	No unacceptable drift, intermittent signal or power collapse.

Deliverable Package

The final project package should include:

• requirements and assumptions;
• installation drawing showing hot side, cold side, heat sink, clamp and sensor location;
• thermoelectric power calculation with derated module temperature difference;
• storage sizing and cold-start energy calculation;
• average and burst power budget;
• piezoelectric charge-amplifier range and ADC resolution calculation;
• sampling, anti-alias and resonance check;
• validation matrix with measured results;
• risk register and corrective actions;
• release recommendation with operating limits.

Risk Register

Risk	Cause	Mitigation
Power margin disappears	Lower machine temperature, poor heat sink, dirty fins or interface aging	Add low-power mode, larger heat sink, better clamp control and minimum-temperature release test.
False low vibration	Sensor detachment, charge amplifier saturation recovery or filtering error	Add data-quality flags, mounted validation and periodic reference check.
Aliasing	Anti-alias filter too high or sampling configuration changed	Lock sampling configuration and verify with injected tones.
Resonance contamination	Sensor mount shifts resonance into diagnostic band	Validate mounted response and control mounting torque or adhesive procedure.
Firmware restart loop	Storage droops during radio burst	Brownout-safe state machine and minimum-storage threshold before radio transmission.
Thermal stress damage	Module clamped across uneven surfaces or exposed to cycling	Use compliant interface, flatness check and thermal-cycle validation.

Release Decision

The prototype is ready for field trial if:

1. harvested average power exceeds the measured node average by the required margin at the minimum operating thermal gradient;
2. storage supports cold start and at least one missed-harvest cycle without data corruption;
3. the piezoelectric channel records normal vibration and startup shock without clipping;
4. noise and quantization are small enough for the alarm or diagnostic threshold;
5. mounted resonance and anti-alias tests support the 10 Hz to 5 kHz diagnostic band;
6. firmware flags invalid power, overload, clipping and missing-sample conditions;
7. the installation record captures clamp force, thermal interface, sensor mount, cable routing and calibration constants.

Do not release the node from calculations alone. The project succeeds only when the installed thermal, mechanical, electrical and firmware evidence all support the same operating envelope.