Build a Universal Testing Machine with Arduino: Load Cell, HX711, Calibration and Stress-Strain Testing

A universal testing machine (UTM) applies a controlled mechanical load to a specimen and measures how the specimen responds. Commercial UTMs can perform tensile, compression, bend, peel, puncture, shear and other mechanical tests. A small Arduino-based machine follows the same basic measurement idea, but with much simpler hardware and lower force capability.

The useful engineering challenge is not simply making a motor pull a sample until it breaks. The machine must measure force, measure or estimate displacement, keep the loading path aligned, move at a controlled rate, log synchronized data, and remain safe when a specimen suddenly fractures.

What does a universal testing machine measure?

At its simplest, a UTM answers two questions at the same time: how much force is being applied, and how much the specimen is deforming (شکل میں تبدیلی). From those measurements, and from the original specimen dimensions, you can calculate useful engineering quantities such as stress, strain, stiffness and ultimate tensile strength.

Measurement	What it tells you	Typical sensor / source
Force	How strongly the machine is pulling or pushing the specimen	Load cell + instrumentation ADC such as HX711
Crosshead position	How far the moving grip or platen has travelled	Stepper count, encoder, linear potentiometer or linear scale
Specimen strain	How much the defined gauge length of the specimen changes	Extensometer or optical / independent displacement measurement
Time / test speed	How fast the test is being performed	Arduino timer + commanded or measured motion
Specimen dimensions	Original cross-sectional area and gauge length	Caliper / micrometer + test setup data

Typical Arduino UTM architecture

The exact hardware varies from one build to another, but a practical low-force Arduino UTM usually contains the following functional blocks. Treat this as a recommended architecture, not a claim that every item below is visible in the ElectroCareHub video.

Subsystem	Typical hardware	Job
Load frame	Rigid single- or dual-column structure	Carries reaction force without excessive bending or twisting
Actuator	Stepper motor + driver + lead screw / ball screw	Moves the crosshead at a controlled speed
Guidance	Linear rods, bearings or rails	Keeps crosshead movement straight and reduces side load
Force sensor	S-type, beam or other suitable load cell	Converts tension/compression into a tiny bridge-voltage change
Signal conditioning	HX711 or another instrumentation ADC	Amplifies and digitizes the load-cell bridge signal
Controller	Arduino Uno or similar board	Reads sensors, commands motion, handles limits and logs data
Displacement sensing	Encoder, linear potentiometer, digital scale or extensometer	Measures movement / specimen elongation
Specimen fixtures	Tensile grips or compression platens	Transfers load into the specimen
Safety	Limit switches, guards, E-stop, current/force limits	Stops dangerous overtravel or overload

How the load cell and HX711 measure force

Most small electronic load cells use strain gauges connected as a Wheatstone bridge. When the load cell bends or stretches slightly, the resistance of the gauges changes and the bridge produces a very small differential voltage. The HX711 is designed to amplify and digitize signals from load cells and strain gauges, which is why it appears so often in Arduino weighing and force-measurement projects.

The HX711 is commonly described as a 24-bit ADC. That is the converter word length, not a promise that your completed machine has 24-bit force accuracy. Real resolution is limited by mechanical vibration, electrical noise, load-cell sensitivity, mounting, temperature, drift and the quality of calibration. Adafruit documents selectable output rates of 10 or 80 samples per second on its HX711 implementation, while SparkFun shows the standard bridge-to-HX711-to-microcontroller connection.

Choose the load cell for the expected force

• Select a capacity comfortably above the highest force you expect, including transient overload, but do not choose a sensor ten or one hundred times larger than necessary if you want good low-force resolution.
• For tension testing, an S-type load cell is convenient because it can be mounted inline and many S-type designs can measure both tension and compression.
• A bar-type load cell must be mounted in the orientation intended by its manufacturer. Clamping both ends incorrectly can make the reading nonlinear or simply wrong.
• Keep the force path centered through the load cell. Side load and bending moments reduce accuracy and can damage the sensor.
• Use mechanical stops or software force limits so a broken specimen or runaway motor does not drive the load cell beyond its rated capacity.

How to calibrate the force measurement

Calibration (کیلیبریشن) converts raw HX711 counts into a force value. A good calibration is not one random weight followed by a convenient multiplication factor. Use several known loads across the part of the range you actually intend to use.

1. Warm up the electronics briefly, leave the machine unloaded and tare the reading to establish zero.
2. Apply a known mass or reference force through the same mechanical load path used during a real test.
3. Record the raw reading after it settles. Repeat at several points, for example near 10%, 25%, 50%, 75% and 90% of the intended working range when safe and practical.
4. Convert known mass to force when reporting in newtons: F = m × g. For ordinary workshop calculations, g ≈ 9.81 m/s².
5. Fit / calculate the calibration factor and then verify it with at least one different reference load that was not used to create the factor.
6. Load and unload the sensor to check hysteresis. Repeat the sequence to judge repeatability (بار بار ایک جیسے نتائج).
7. Save the calibration factor together with the load-cell serial number / capacity and the date. Recalibrate after mechanical changes, overloads or unexplained drift.

Calibration check	Why it matters	What a problem looks like
Zero stability	Confirms tare and drift are under control	Reading slowly walks while nothing is loaded
Multi-point linearity	Checks scale factor across the range	Correct at one weight, wrong at higher loads
Repeatability	Shows whether setup gives consistent results	Same reference load gives noticeably different readings
Hysteresis	Compares loading vs unloading	Reading depends on direction of force change
Off-axis sensitivity	Reveals mounting/alignment issues	Reading changes when the fixture is nudged sideways

Measuring displacement: stepper travel vs real specimen strain

A stepper-driven lead screw gives a convenient estimate of crosshead motion. If the lead screw moves P millimetres per revolution, the motor has S full steps per revolution, and the driver uses M microsteps per full step, then the nominal motion per microstep is:

distance per microstep = P / (S × M)

That value describes commanded machine movement. It does not automatically describe the change in specimen gauge length. Backlash, lead-screw error, elastic deflection of the frame, grip deformation and slip all sit between the motor command and the specimen.

Method	Advantages	Limitations
Stepper count only	Very cheap; no extra sensor; useful for coarse crosshead position	Does not detect missed steps; includes backlash, frame compliance and grip effects
Rotary encoder	Can confirm screw / motor rotation	Still measures drive motion rather than specimen deformation
Linear potentiometer / encoder	Measures crosshead or fixture travel directly	Still includes machine / grip compliance unless attached across specimen gauge length
Clip-on extensometer	Direct gauge-length strain measurement; good for modulus/yield work	More complex and must suit specimen travel and force environment
Optical / video extensometer	No contact with specimen; useful for fragile/high-elongation samples	Requires good optics, calibration, markers and image processing

From force and elongation to stress and strain

Force and movement are only the raw measurements. Material properties depend on the original specimen geometry.

Quantity	Equation	Meaning
Engineering stress, σ	σ = F / A₀	Applied force divided by original cross-sectional area
Engineering strain, ε	ε = ΔL / L₀	Change in gauge length divided by original gauge length
Young’s modulus, E	E ≈ Δσ / Δε in the initial linear region	Material stiffness when the test and strain measurement are suitable
Ultimate tensile strength	Maximum engineering stress reached	Highest engineering stress before/at necking or fracture
Elongation at break	Final extension relative to original gauge length	A measure of ductility, when measured by the specified method

If force is in newtons and the original cross-sectional area is in square millimetres, stress comes out in N/mm², which is numerically equal to megapascals (MPa). This makes workshop calculations convenient, provided the original width/thickness or diameter is measured carefully.

A practical tensile-test workflow

1. Choose a specimen and test method appropriate to the educational goal. Measure its original width, thickness / diameter and gauge length before loading.
2. Inspect the grips, load cell, frame and fasteners. Set software force and travel limits below the weakest safe component rating.
3. Install the specimen so its axis lines up with the load cell and crosshead. Poor alignment introduces bending and can cause premature grip-end failure.
4. Bring the grips into light contact / minimal preload as appropriate, then tare the force channel. Do not use a large preload just to make the graph start neatly.
5. Zero or record the displacement / extensometer reading at the defined starting gauge length.
6. Start data logging before motion. Record time, force and displacement from the same test clock.
7. Move the crosshead at the planned test speed. Stop automatically at the chosen force, displacement, fracture detection or safety limit.
8. Keep hands away from the specimen and drive system while the test is running. Brittle fragments and snapped specimens can become projectiles.
9. After the test, save raw data first. Calculate stress/strain from a copy so the original measurements remain available.
10. Repeat the test on multiple specimens if you want to discuss repeatability rather than one dramatic break.

What data should the Arduino log?

A useful UTM log should preserve raw measurements and calculated values. Do not save only the final tensile-strength number; that destroys most of the diagnostic value of the test.

Field	Example unit	Why keep it
Timestamp / elapsed time	ms or s	Lets you calculate test rate and align sensors
Raw HX711 count	counts	Allows later recalibration / debugging
Force	N	Primary load measurement
Crosshead position	mm	Machine movement and limit tracking
Independent extension	mm	Preferred input for specimen strain when available
Specimen area A₀	mm²	Required for engineering stress
Gauge length L₀	mm	Required for engineering strain
Stress	MPa	Material-response calculation
Strain	dimensionless or %	Material deformation calculation
Machine state	IDLE / RUN / STOP / BREAK	Explains where each sample belongs in the test sequence

Arduino software architecture

The control code should keep measurement, motion and safety separate enough that one slow function does not hide a dangerous condition. A sensible loop looks like this:

1. Initialize HX711, position sensor, motor driver, limit switches and serial / SD logging.
2. Wait for the operator to confirm the fixture is safe, then tare force and zero position.
3. Read force and position at a consistent rate; apply only documented filtering.
4. Check hard limits first: E-stop, limit switches, maximum force, maximum travel and sensor faults.
5. Command the next motor step / velocity only if all limits are healthy.
6. Write synchronized data to serial, SD card or a connected computer.
7. Detect end-of-test conditions such as fracture or target displacement and stop motion safely.
8. Require an explicit reset before allowing another move after a safety trip.

Noise, filtering and sampling rate

A UTM is a dynamic measurement system. Heavy averaging can make a graph look smooth while hiding a real peak load or delaying fracture detection. Too little filtering can turn vibration and electrical pickup into fake “material behavior.”

• Mount the HX711 close enough to the load cell to keep the millivolt-level bridge wiring short, and use shielded / twisted wiring where appropriate.
• Separate stepper-motor power wiring from load-cell signal wiring. Motor drivers are noisy switching devices.
• Choose a sample rate appropriate to test speed. A very slow tensile test may be fine at low sample rate; a fast fracture event can be poorly represented if the data rate is too low.
• Record raw or lightly processed samples when possible, then apply clearly documented analysis filtering later.
• If using averaging, store the method and number of samples so results can be reproduced.
• Do not infer more decimal places than the machine can physically resolve. Display formatting is not measurement accuracy.

Where DIY UTM results commonly go wrong

Problem	Likely cause	Corrective action
Force reading drifts at zero	Temperature drift, mechanical preload, noisy supply, sensor creep	Warm up, remove preload, improve wiring, retare and check load-cell mounting
Reading is negative when pulling	Signal polarity reversed	Reverse signal leads or change sign in software after confirming wiring
Correct at one calibration weight, wrong elsewhere	Poor calibration, nonlinear mounting, overload or mechanical binding	Use multi-point calibration; inspect load path and sensor mounting
Graph is very noisy when motor runs	EMI / ground coupling from stepper driver	Separate power/signal wiring, improve grounding/shielding, decouple supplies
Arduino resets during motion	Motor supply sag or electrical noise	Use proper motor PSU/driver, grounding and decoupling; do not power motor from Arduino rail
Crosshead moves but measured extension is wrong	Backlash, missed steps, frame compliance or grip slip	Add position feedback / extensometer and improve mechanics
Specimen always breaks at grip	Grip damage, stress concentration or poor alignment	Use correct grip geometry, alignment and specimen preparation
Calculated strength is impossibly high/low	Wrong area units, wrong force units or specimen dimensions	Check N vs kgf and mm² vs m²; remeasure original section
Peak force changes strongly between repeats	Misalignment, inconsistent specimens, speed variation or grip slip	Control specimen geometry, alignment and test speed; repeat calibration
HX711 reading freezes / becomes erratic	Wiring, library timing, electrical noise or sensor fault	Check DAT/CLK, supply, grounding and code flow; verify with static load

Safety: a small machine can still injure you

• Guard the lead screw, coupler and moving crosshead. Pinch points can trap fingers even at modest motor torque.
• Use physical travel limit switches. Software limits alone do not protect against a crashed program, disconnected sensor or wrong calibration.
• Add an emergency stop that removes actuator power, not just a software button on the Arduino screen.
• Wear eye protection and use a clear shield when testing brittle plastics, composites, wires or parts that can snap suddenly.
• Do not stand in line with a tensioned specimen. Stored elastic energy is released when it breaks.
• Rate grips, bolts, frame, lead screw and load cell above the planned working load with an appropriate engineering safety margin.
• Stop immediately if the frame bends, fasteners loosen, the load cell goes off-scale or the machine makes abnormal mechanical noise.

DIY Arduino UTM vs commercial universal testing machine

Area	Arduino DIY machine	Commercial / standards-oriented UTM
Primary value	Learning, prototyping, comparative experiments	Traceable engineering and quality-control measurements
Force capacity	Usually low to moderate and build-dependent	Specified and verified across rated range
Frame stiffness / alignment	Depends heavily on fabrication quality	Engineered and characterized for test accuracy
Strain measurement	Often crosshead estimate unless extra sensor added	Dedicated extensometers / validated strain systems commonly used
Speed control	Stepper-command based; may lack closed-loop verification	Controlled and documented test rates / feedback
Calibration	User-created with masses or reference loads	Traceable calibration / verification procedures
Safety	Depends on builder discipline	Purpose-designed guards, limits and safety systems
Standards compliance	Do not assume	Can be configured / verified for specific standards
Cost	Low	Much higher

Frequently asked questions

Can Arduino make a real universal testing machine?

Arduino can control a useful low-cost materials-testing rig, but the quality of the machine depends more on mechanics, sensing, calibration, alignment, displacement measurement and safety than on the microcontroller itself.

Why use HX711 with a load cell?

A strain-gauge load cell produces a very small differential signal. HX711 provides high-gain differential conversion designed for load-cell / weigh-scale applications and is easy to interface with a microcontroller.

Should force be shown in kg or newtons?

Use newtons for force. Kilograms describe mass. If calibration uses known masses, convert the applied weight to force before reporting engineering results.

Can I calculate tensile strength from load-cell force?

Yes, if you know the maximum force and the correct original specimen cross-sectional area: engineering tensile strength = maximum force / original area. The result is only as good as the force, area and test setup.

Can I calculate strain from stepper motor steps?

You can estimate crosshead displacement, but it is not the same as direct specimen strain. Frame flex, backlash, grip slip and fixture seating introduce error. Use an extensometer or independent gauge-length displacement measurement for serious strain work.

What load cell should I use?

Choose a type suitable for tension/compression as required and a capacity modestly above the expected maximum load. Also consider overload rating, mounting, sensitivity and available calibration references.

Why are my HX711 readings noisy?

Common causes are motor-driver EMI, long unshielded bridge wiring, poor grounding, vibration, unstable supply voltage, excessive averaging settings or an improperly mounted load cell.

How do I calibrate the UTM?

Tare at zero, apply multiple known reference loads through the real load path, calculate the conversion factor, verify with separate reference points, and check repeatability and hysteresis.

Can this machine meet ASTM E8/E8M?

Not automatically. ASTM tension testing depends on specimen geometry, machine and force verification, test rate, alignment, strain measurement and other controlled requirements. A DIY machine should be described as educational unless it has been formally verified for the standard.

Can the same machine do compression tests?

Potentially, if the frame, actuator, load cell and fixtures are designed for compression and correct platens are fitted. Never assume a tension grip or sensor mounting is safe in compression.

How many samples should I test?

One specimen is useful for demonstrating the machine, but it says little about repeatability. For comparisons, test multiple specimens prepared and loaded the same way and report the spread, not only the best-looking curve.

What should I upgrade first for better data?

After safe mechanics and reliable force calibration, add a real displacement / extensometer measurement, improve alignment and frame stiffness, and log synchronized raw data. Those upgrades usually matter more than changing to a faster microcontroller.