Size control and online intelligent detection technology of precision spring
Dimensional Control and Automatic On-line Inspection System in Precision Spring Manufacturing
Introduction
In the field of miniature precision springs (wire diameter 0.1~ 2.0 mm) and high-stress automotive springs, the dispersion between size and force value directly determines the quality and functional consistency of the system assembly. Taking the shift spring in an automatic transmission as an example, if the free length tolerance exceeds ±0 mm, it may lead to abnormal shifting force, cause frustration and even gear failure; if the stiffness tolerance of the safety valve spring exceeds ±5%, it may cause the valve opening pressure to deviate from the design value, causing equipment overpressure accidents.
Traditional production methods rely on manual sampling and offline testing, which has problems such as lag, incomplete coverage, and data traceability. In 2025, industry leaders have widely deployed online 100% full inspection systems, combined with the closed-loop control of CNC spring coiling machines, to increase the process capability index Cpk from 0.8 to more than 1.33, and the failure rate from 5000 ppm to less than 100 ppm.
This paper systematically introduces the dimensional tolerance standard of precision spring, the precision control technology of CNC coil spring, the principle of on-line testing equipment and the application practice of statistical process control (SPC).
First, the main geometric and mechanical parameters of precision springs
1.1 Key dimensional parameters
Parameter definition Typical tolerances (precision grade)
Wire diameter (d) Wire diameter ±0 mm (fine spring ±0.002 mm)
Outer Diameter (D_e)/Inner Diameter (D_i) Coil Outer/Inner Diameter ±0 mm~ ±0.1 mm
Free length (L_0) Length without load ±0 mm (miniature spring ±0.03 mm)
Total number of turns (N_t) Effective number of turns + number of bearing turns No tolerance, but the number of turns error ≤ 0.1 turns
Verticality The vertical deviation between the two ends and the axis ≤ 0.5 or ≤ 05L_0
Pitch Uniformity Maximum Difference of Adjacent Effective Circle Pitch ≤ 0.05 mm
1.2 Key mechanical parameters
Stiffness (k): the force generated by the unit deformation, unit N/mm. Tolerance generally requires ±5%~ ±10%.
Load at specified height: e.g. force value F at L = 20 mm. Tolerance is usually ±5%.
Permanent deformation: spring compression to the maximum working stroke after recovery, free length change. Requirements ≤ 0.5% L_0.
Load loss rate: the proportion of force value attenuation after high temperature or cycle. Valve spring requirements ≤ 3%.
Second, the precision control technology of CNC spring coiling machine
Modern CNC spring coiling machine is composed of wire feeding wheel, variable diameter mechanism, pitch control cam, and cutting knife, which are independently driven by servo motors. The core link of precision control:
2.1 Line feeding accuracy
The pressing force of the wire feed wheel forms a closed loop with the encoder feedback. Factors affecting the wire feed error include:
The lubrication state of the steel wire surface (changes can lead to slippage);
Feeding wheel wear (weekly calibration compensation);
Steel wire curvature (pre-straightening required).
The repeated positioning accuracy of the high-precision model can reach ±0 mm, corresponding to the spring with a wire diameter of 1.0 mm, and the free length error can be controlled at ±0 mm.
2.2 Variable diameter control
The winding diameter is controlled by cam or skateboard. The latest technology adopts real-time detection of wire diameter + dynamic compensation of variable diameter: a laser caliper is installed at the winding outlet, and the measured outer diameter is fed back to the controller to adjust the position of the variable diameter cam in real time. The outer diameter tolerance is ±0 mm.
2.3 Pitch control
The pitch is controlled by pitch cam or servo lever. For precision springs, optical online detection of spring wire diameter + pitch is often used to judge whether the gap between adjacent rings is uniform through machine vision. If it exceeds the tolerance, it will be alarmed or automatically adjusted.
2.4 Cutting accuracy
Poor coordination between the cutter and the mandrel can cause end face burrs or excessive length. Advanced models use servo rotary cutting, and the cutter rotates synchronously with the spring to obtain a flat end face.
III. Composition of a fully automatic online detection system
3.1 Optical dimension detection
Equipment principle: high-resolution CMOS camera + backlight + edge extraction algorithm.
Test items: free length, outer diameter/inner diameter, pitch, end face parallelism, verticality.
Detection speed: 60 to 200 pieces per minute (depending on the spring size).
Accuracy: Length measurement accuracy ±0 mm, diameter ±0 mm.
Advantages: non-contact, no deformation, full inspection.
Project case: A valve spring production line is equipped with 4 optical detectors, which are located after the coil spring, after heat treatment, after shot peening, and before the final packaging to achieve full-process size monitoring.
3.2 Force value automatic testing device
Principle: The servo press compresses the spring at a constant speed to a specified height, reads the force value through the force sensor, and compares it with the standard curve.
Test points: Usually test 2 to 4 specified height points (e.g. pre-pressing position, working position, maximum compression position).
Output index: stiffness, specified height force value, permanent deformation.
Repeat accuracy: force value ±0%, displacement ±0 mm.
Full inspection feasibility: the beat can reach 20 to 30 pieces/minute, suitable for 100% online inspection.
3.3 Eddy current flaw detection and surface defect detection
Used to detect small cracks, folds, scratches on the surface of the spring. The vortex probe scans along the surface of the spring, and the impedance change reflects the depth of the defect. It can detect surface opening defects with a depth of ≥ 0.05 mm. Combined with the rotating mechanism, it can cover the entire surface of the spring.
IV. Statistical Process Control (SPC) and Failure Rate Optimization
4.1 Calculation of process capability index Cpk
Cpk = min[ (USL - μ)/(3σ), (μ - LSL)/(3σ) ]
Where USL/LSL is the upper and lower specification limits, μ is the mean, and sigma is the standard deviation.
Industry Benchmark:
Cpk < 0.67 is unacceptable and needs to be improved.
0.67 ≤ Cpk < 1.00 barely qualified, there is a risk of non-conforming products;
1.00 ≤ Cpk < 1.33 Good;
Cpk ≥ 1.33 Excellent with a failure rate of < 66 ppm.
Case: A spring factory conducts SPC monitoring of stiffness, and collects 125 samples: mean μ = 10.02 N/mm, standard deviation (sigma) = 0.12 N/mm, specification is 10.0 ± 0.5 N/mm. Then Cpk = min ((10.5-10)/(30.12), (10.02-9)/(30.12)) = min (1.33, 1.44) = 1.33. Excellent process ability.
4.2 Application of Control Charts
The commonly used Xbar-R plot (mean-range plot) monitors process stability and long-term drift. If 7 consecutive points rise or fall, or the data points exceed the upper and lower control limits, the process is determined to be out of control, and the cause (such as tool wear, material batch changes) needs to be investigated immediately.
4.3 Failure rate optimization actual combat
Problem cause analysis Countermeasure effect
The slip of the free-length over-difference wire feeding wheel increases the pressing force, and the failure rate of regular cleaning of the wheel groove is reduced from 3% to 0.5%.
The tensile strength of materials with large stiffness dispersion batch fluctuation is tested for each batch of incoming materials, and the pre-adjusted coil spring parameter Cpk is increased from 0.9 to 1.2
End face parallelism difference grinding spring fixture wear calibration fixture before each shift, add online parallelism detection parallelism unqualified rate approaches zero
Smart Manufacturing Trends: Digital Twins and AI Screening
5.1 Digital twin closed-loop control
The data of spring coiling machine, heat treatment furnace, shot peening machine and testing equipment are connected to the MES system in real time to establish a digital twin model of the spring production line. When the testing station finds that a certain parameter has a drift trend, the model adjusts the setting value of the front equipment in reverse (such as wire feeding speed, heating temperature) to achieve predictive adjustment and avoid waste.
5.2 Deep learning appearance defect screening
For minor defects on the surface of the spring (pits and rust spots less than 0.1 mm), traditional visual rule algorithms are difficult to stably detect. Now convolutional neural networks (CNN) are used to train the classification model, input the surface image of the spring, and output the pass/fail judgment. After the training dataset contains 100,000 labeled pictures, the model accuracy can reach more than 99.5%.
conclusion
The consistency of the size and force value of the precision spring is no longer the art of "relying on the experience of the master to adjust the machine", but a complete technical system composed of CNC spring coiling machine, online optical inspection, full force value inspection, SPC control and AI vision. Enterprises that implement online 100% inspection can not only control the failure rate within 100 ppm, but also provide traceable inspection data packets for downstream customers, which can significantly improve the quality reputation. The tolerance standards, testing equipment parameters and SPC methods given in this paper can be directly used as a technical reference for spring manufacturers and purchasers.
BQUQ is a professional metal spring manufacturer, please send us drawings, and our company will quote you within 12 hours.
