Electric connector weak light assembly vibration compensation alignment method

Abstract

The invention relates to the technical field of electric connector weak light assembly vibration compensation alignment methods, particularly discloses an electric connector weak light assembly vibration compensation alignment method, and aims to solve the problems of low electric connector assembly positioning precision and high pseudo soldering rate caused by vibration in a weak light environment. The method comprises the following steps: monitoring the thermal radiation change of a vibration source in real time through an infrared thermal imaging sensor, and pre-judging a vibration acceleration vector in combination with a lightweight long-short-term memory neural network; synchronously triggering dual polarization angle imaging to respectively capture a housing reference mark and an enhanced terminal metal feature; the vibration data and the polarization image flow are fused in the field programmable gate array, a self-adaptive optical flow algorithm is adopted to resolve the displacement compensation amount, and a piezoelectric ceramic actuator is driven to carry out submicron order reverse displacement compensation; when the compensation amount exceeds a threshold value, a low-delay correction instruction is sent to the robot through an Ethernet control automation technology bus, and the steady-state error is ensured to be smaller than 0.3 micrometer by combining a strain gauge fine tuning feedback loop. According to the technical scheme, the assembling precision, efficiency and production line adaptability are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of electrical connector assembly technology, and specifically relates to a method for vibration compensation alignment during low-light assembly of electrical connectors. Background Technology

[0002] With the rapid development of high-end electronic manufacturing towards miniaturization and high density, the automated precision assembly of electrical connectors has become a crucial link in ensuring signal integrity and product reliability. In this field, the sub-micron alignment accuracy between terminals and housings directly determines contact resistance, crosstalk levels, and long-term service stability. Currently, mainstream production lines generally rely on machine vision-based closed-loop positioning systems, which use optical imaging to identify reference features and guide the actuator to complete the mating operation. However, in actual industrial scenarios, the coupling effect of low-light conditions and high-frequency mechanical vibrations severely restricts the robustness and real-time performance of existing systems, especially in the assembly of miniature connectors with a pitch of less than 0.5 mm, where traditional solutions struggle to meet alignment tolerance requirements within ±3 μm.

[0003] Among these challenges, precise alignment technology for electrical connectors in low-light assembly environments faces three core difficulties. First, existing technologies suffer from a sharp drop in the signal-to-noise ratio of CMOS sensors under low illumination (<50 lux), leading to the loss of diffuse reflection features on metal terminal surfaces, failure to identify markings on plastic shells, and a significant increase in feature point extraction errors. Second, high-frequency disturbances of 50–200Hz generated by vibration sources such as conveyor belt motors and pneumatic clamps in the production line are amplified by the transmission chain, causing resonance in the clamping mechanism. Passive damping methods relying solely on rubber dampers cannot effectively suppress such dynamic deviations. Third, and more critically, mainstream image processing workflows rely on CPU-based software-level optical flow algorithms, resulting in a delay of over 1.2ms from acquisition to compensation command output, far exceeding the 200μs real-time control window. This causes phase lag or even negative interference, further exacerbating positioning inaccuracies.

[0004] Therefore, there is an urgent need for a dynamic alignment method that can achieve full-link hardware acceleration of perception, decision-making and execution under conditions of coexistence of low light and strong vibration, so as to break through the technical bottleneck of existing single-modal vision or single vibration suppression strategies and solve the problems of real-time performance, robustness and adaptability in high-precision assembly of micro electrical connectors.

[0005] In view of this, this application proposes a vibration compensation alignment method for low-light assembly of electrical connectors. Summary of the Invention

[0006] The purpose of this invention is to provide a method for assembling electrical connectors under low light conditions with vibration compensation, which can effectively solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for aligning electrical connectors during low-light assembly with vibration compensation includes the following specific steps: Step 1: Monitor the changes in thermal radiation of the vibration source in real time using an infrared thermal imaging sensor. Collect thermal radiation data at a frame rate of 120 frames per second, focusing on the motor bearing area. When the bearing temperature rise exceeds 2 degrees Celsius per second, it is determined that the vibration source is activated. Output the vibration acceleration vector through the lightweight long short-term memory neural network model built into the field programmable gate array. Step 2: Synchronously trigger the dual polarization angle imaging of the polarization vision system. The polarization angle switching timing is precisely controlled by the field programmable gate array clock management unit. At 60 microseconds, the 0-degree polarization angle is triggered to capture the reference mark on the shell, and at 65 microseconds, the polarization angle is switched to 90 degrees to enhance the features of the terminal metal surface. Step 3: In the field-programmable gate array (FPGA) processing unit, the vibration acceleration vector and polarization image flow are fused, and the displacement compensation is calculated using an adaptive optical flow algorithm. The vibration acceleration integral term accounts for 60% of the displacement compensation weight, and the image optical flow term accounts for 40% of the weight. Step 4: Generate a pulse width modulation control signal based on the displacement compensation amount to drive the piezoelectric ceramic actuator to push the terminal clamping mechanism to move in the opposite direction. The piezoelectric ceramic actuator has a resolution of 0.1 micrometers and a stroke of ±50 micrometers. Step 5: When the absolute value of the displacement compensation is greater than the set value, which can be 5 micrometers, a motion correction command is sent to the assembly robot via the Ethernet control automation technology bus. The delay of the motion correction command is less than 50 microseconds. Step 6: A fine-tuning feedback loop is formed by using built-in strain gauges to provide real-time feedback on displacement errors and trigger the fine-tuning cycle, ensuring that the compensation accuracy reaches ±0.1 micrometers.

[0008] Preferably, the infrared thermal imaging sensor is installed 15 cm directly above the assembly head, at a 45-degree angle to the motor axis, to avoid direct thermal radiation interference. The collected thermal radiation data is an 8-bit grayscale image sequence of 16 x 16 pixels per frame.

[0009] Preferably, the lightweight long short-term memory neural network model has fewer than 10,000 parameters, takes 10 consecutive frames of thermal radiation image sequence as input, and outputs a three-dimensional vibration acceleration vector, achieving a measured prediction accuracy of 96.2%.

[0010] Preferably, the polarized light vision system adopts a ring layout, includes 8 groups of 850 nanometer light-emitting diodes, the polarization angle can be switched between 0 degrees and 90 degrees, and is coaxially mounted with the Sony image sensor, 20 cm away from the terminal contact surface.

[0011] Preferably, during the dual polarization angle imaging process, the time difference between the acquisition of two frames is less than 5 microseconds. The 0-degree polarization angle imaging is used to suppress ambient light interference, and the 90-degree polarization angle imaging enhances the reflection characteristics of the terminal metal surface through Brewster angle.

[0012] Preferably, the field-programmable gate array uses Xilinx VLSI devices with 1,123,600 logic cells and a block random access memory capacity of 62,400 kilobits. It processes infrared prediction data and polarization image streams in parallel through an advanced scalable interface stream bus.

[0013] Preferably, the adaptive optical flow algorithm is embedded as a hardware intellectual property core in a field-programmable gate array, adopts the Horn-Schuck optical flow calculation principle, and performs gradient calculation and integral operation in parallel, completing the processing of 1280 by 1024 pixels in the entire area within 140 microseconds.

[0014] Preferably, the pulse width modulation control signal is output through a low-voltage differential signal interface, the conversion coefficient between voltage and displacement compensation is 0.2 volts per micrometer, and the signal bit width is 12 bits.

[0015] Preferably, the piezoelectric ceramic actuator is rigidly integrated into the bottom of the clamping mechanism, with built-in strain gauges to monitor displacement error in real time. When the error exceeds a set threshold, a fine-tuning cycle is automatically triggered to ensure that the steady-state error is less than 0.3 micrometers.

[0016] Preferably, the Ethernet control automation technology bus transmission protocol ensures that the motion correction command delay is less than 50 microseconds, and immediately activates the safety protocol to send the correction command to the assembly robot when the absolute value of the displacement compensation is greater than 5 micrometers.

[0017] Preferably, the fine-tuning feedback loop adopts a closed-loop control strategy, which collects displacement data in real time through strain gauges, compares it with the preset target value, and generates a correction signal to continuously optimize the compensation accuracy.

[0018] Preferably, the method operates under low-light conditions with an illumination level below 50 lux, and ensures that the positioning error does not exceed ±3 micrometers through a dual-modal fusion mechanism. The actual measured positioning error is 2.7 micrometers under a 40 lux environment.

[0019] Preferably, the entire process takes less than 200 microseconds, with the vibration prediction stage taking 60 microseconds, the polarization light feature capture stage taking 80 microseconds, the optical flow field fusion calculation stage taking 80 microseconds, and the compensation command generation stage taking 60 microseconds.

[0020] Preferably, the method supports a full range of connectors with terminal pitches from 0.2 to 1.0 mm, and does not require retraining the model when switching production lines, significantly improving production line adaptability.

[0021] Preferably, the total power consumption of the compensation process is less than 8 watts, which is 92% lower than that of the graphics processor solution, and there is no local temperature rise problem, ensuring long-term stable operation of the system.

[0022] Preferably, the method was tested on a miniature electrical connector assembly line for 30 consecutive days, showing that the positioning error was improved by 73.8% compared to the traditional solution, the signal crosstalk was improved to -58 dB, and the cold solder joint rate was reduced to 0.9%.

[0023] Preferably, the field-programmable gate array (FPGA) processing unit further includes a clock management module to precisely control the timing of each stage and ensure strict synchronization of data acquisition, processing and output.

[0024] Preferably, the polarized light vision system employs an adaptive histogram equalization preprocessing algorithm to perform differentiated enhancement processing on images with different polarization angles. For images with a 0-degree polarization angle, median filtering is used to suppress shell noise, and for images with a 90-degree polarization angle, gamma correction is used to enhance metal reflection features.

[0025] Preferably, in the process of calculating the displacement compensation, the vibration acceleration integral term is obtained by numerical integration, and the image optical flow term is obtained by calculating the image gradient. The two are fused together with a weight of 6:4 to generate the final displacement compensation.

[0026] Preferably, the response characteristics of the piezoelectric ceramic actuator are precisely calibrated, with a rise time of 85 microseconds and an overshoot of less than 5%, ensuring fast and accurate displacement compensation.

[0027] Preferably, the method also includes a self-test function to periodically verify the accuracy of the sensors and the performance of the actuators, ensuring the reliability of the system's long-term operation.

[0028] Compared with the prior art, the present invention has the following beneficial effects: By employing a dual-modal fusion mechanism of infrared thermal imaging prediction and polarization-based visual correction, combined with field-programmable gate array (FPGA) hardware acceleration, a 200-microsecond dynamic compensation closed loop is achieved. In low-light environments, the positioning error is compressed to 2.7 micrometers, a 73.8% improvement over traditional solutions. Simultaneously, signal crosstalk is reduced to -58 dB, and the cold solder joint rate is reduced to 0.9%. It supports a full range of connectors, and production line switching does not require model retraining. The power consumption throughout the compensation process is less than 8 watts, with no local temperature rise issues, significantly improving the accuracy, efficiency, and reliability of electrical connector assembly. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall technical solution architecture of the low-light assembly vibration compensation alignment method for electrical connectors proposed in this invention. Figure 2 This is a schematic diagram of the core principle framework of the dual-modal fusion mechanism of infrared thermal imaging prediction and polarized light visual correction in this invention; Figure 3 This is a flowchart illustrating the logical process of adaptively fusing vibration acceleration vector and polarization image stream to calculate displacement compensation in this invention. Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow of the piezoelectric ceramic actuator drive and fine-tuning feedback closed-loop control in this invention; Figure 5 This is a key stage processing framework diagram for the generation of timing synchronization and motion correction instructions throughout the entire process in this invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0031] Example 1: Currently, with the rapid development of high-end electronic manufacturing towards miniaturization and high density, the automated precision assembly of electrical connectors has become a key link in ensuring signal integrity and product reliability. In this field, the sub-micron alignment accuracy of terminals and housings directly determines contact resistance, crosstalk level, and long-term service stability. Current mainstream production lines generally rely on machine vision-based closed-loop positioning systems, which use optical imaging to identify reference features and guide the actuator to complete the mating operation. However, in actual industrial scenarios, the coupling effect of weak lighting conditions and high-frequency mechanical vibration severely restricts the robustness and real-time performance of existing systems, especially in the assembly of micro-connectors with a pitch of less than 0.5 mm, where traditional solutions struggle to meet alignment tolerance requirements within ±3 microns. To address the aforementioned technical issues, this invention proposes a dual-modal fusion mechanism combining infrared thermal imaging prediction and polarized light visual correction. This mechanism, coupled with field-programmable gate array (FPGA) hardware acceleration, achieves a 200-microsecond-level dynamic compensation closed loop. In low-light environments, the positioning error is reduced to 2.7 micrometers, a 73.8% improvement over traditional solutions. Simultaneously, signal crosstalk is reduced to -58 dB, and the solder joint failure rate is lowered to 0.9%. This invention supports all connector series and eliminates the need for model retraining during production line switching. The entire compensation process consumes less than 8 watts, with no localized temperature rise issues. This significantly improves the accuracy, efficiency, and reliability of electrical connector assembly and is applied to vibration compensation alignment methods for low-light assembly of electrical connectors.

[0032] refer to Figure 1The overall technical architecture of the low-light assembly vibration compensation alignment method for electrical connectors proposed in this invention includes an infrared thermal imaging sensor, a polarized light vision system, a field-programmable gate array (FPGA) processing unit, a piezoelectric ceramic actuator, a strain gauge feedback module, and an Ethernet control automation bus communication interface. Specifically, the infrared thermal imaging sensor is used to monitor changes in thermal radiation from the vibration source in real time; the polarized light vision system is used to capture the reference marks on the outer shell and the surface features of the terminal metal under low-light conditions; the FPGA processing unit serves as the core computing platform, integrating a clock management module, a lightweight long short-term memory neural network model, an adaptive optical flow algorithm hardware intellectual property core, and data fusion logic; the piezoelectric ceramic actuator is used to perform high-resolution displacement compensation; the strain gauge forms a fine-tuning feedback loop to ensure steady-state accuracy; and the Ethernet control automation bus is used to collaboratively correct the motion trajectory with an external assembly robot in large displacement scenarios.

[0033] In the aforementioned weak-light assembly vibration compensation alignment method for electrical connectors, step (1) involves real-time monitoring of the thermal radiation changes of the vibration source using an infrared thermal imaging sensor. Thermal radiation data is collected at a frame rate of 120 frames per second, with a focus on the motor bearing area. When the bearing temperature rise exceeds 2 degrees Celsius per second, the vibration source is considered activated. The vibration acceleration vector is output through the lightweight long short-term memory neural network model built into the field-programmable gate array (FPGA). Specifically, the infrared thermal imaging sensor is installed 15 cm directly above the assembly head, at a 45-degree angle to the motor axis to avoid direct thermal radiation interference. The collected thermal radiation data is a sequence of 16 x 16 pixels, 8-bit grayscale images per frame. The sensor uses an uncooled microbolometer array with a response band of 8 to 14 micrometers, a thermal sensitivity better than 50 milliklvin, and an inter-frame time interval of 8.33 milliseconds. The lightweight long short-term memory neural network model in the FPGA processing unit has fewer than 10,000 parameters. The input is a continuous sequence of 10 frames of thermal radiation images, and the output is a three-dimensional vibration acceleration vector. The measured prediction accuracy reaches 96.2%. The model was trained offline, with the training dataset containing motor operating states under different loads, speeds, and ambient temperatures, covering normal operating conditions and typical fault modes. Fixed-point quantization was used during model deployment, converting floating-point weights to 8-bit integers to accommodate the logic resource constraints of the field-programmable gate array (FPGA). The vibration source activation determination logic is based on the temperature rise slope calculation within a sliding window: linear fitting is performed on the grayscale values ​​of the center pixels of the bearing region in five consecutive frames of images. If the temperature rise rate corresponding to the slope is greater than 2 degrees Celsius per second, the prediction process is triggered. This stage is strictly controlled within 60 microseconds to ensure that the prediction is completed before the vibration disturbance propagates to the clamping mechanism.

[0034] In the above-mentioned weak light assembly vibration compensation alignment method for electrical connectors, step (2) synchronously triggers the dual polarization angle imaging of the polarized light vision system. The polarization angle switching sequence is precisely controlled by the field-programmable gate array clock management unit. At 60 microseconds, the 0-degree polarization angle is triggered to capture the reference mark on the outer shell, and at 65 microseconds, it switches to a 90-degree polarization angle to enhance the characteristics of the terminal metal surface. Specifically, the polarized light vision system adopts a ring layout, containing 8 groups of 850-nanometer light-emitting diodes. The polarization angle can be switched between 0 degrees and 90 degrees. It is coaxially mounted with the Sony image sensor and is 20 cm away from the terminal contact surface. The time difference between the acquisition of two frames of images is less than 5 microseconds. The 0-degree polarization angle imaging is used to suppress ambient light interference, and the 90-degree polarization angle imaging enhances the reflection characteristics of the terminal metal surface through Brewster angle. The clock management module of the field-programmable gate array generates a precise trigger pulse sequence to control the synchronous operation of the light-emitting diode driving circuit and the polarizer rotation mechanism (or liquid crystal polarization modulator). At a 0-degree polarization angle, the linearly polarized light emitted from the light source remains essentially unchanged after diffuse reflection from the plastic surface of the outer casing. This allows for efficient reception by the coaxial image sensor, clearly displaying printed or laser-etched reference marks. At a 90-degree polarization angle, the incident light's polarization direction is perpendicular to the normal plane of the metal terminal surface. According to Fresnel's equations, near the Brewster angle (approximately 87 degrees for copper alloys), the reflectivity of the s-polarization component is significantly higher than that of the p-polarization component, thus enhancing the contrast between the metal edge and the background. The image sensor employs a global shutter CMOS device with a resolution of 1280 x 1024 pixels, a pixel size of 3.45 micrometers, and a maximum frame rate of 200 frames per second. After acquisition, the system immediately applies an adaptive histogram equalization preprocessing algorithm to both images: the 0-degree polarization angle image uses median filtering to suppress casing noise, with a filtering window of 3 x 3 pixels; the 90-degree polarization angle image uses gamma correction (γ = 0.6) to enhance metal reflection characteristics and improve the signal-to-noise ratio. The total time for this stage is 80 microseconds, of which 65 microseconds are for illumination switching and image acquisition, and 15 microseconds are for preprocessing.

[0035] In the above-mentioned weak light assembly vibration compensation alignment method for electrical connectors, step (3) involves fusing the vibration acceleration vector and polarization image stream in the field-programmable gate array (FPGA) processing unit, and using an adaptive optical flow algorithm to calculate the displacement compensation amount. The vibration acceleration integral term accounts for 60% of the displacement compensation amount, and the image optical flow term accounts for 40%. Specifically, the FPGA uses Xilinx VLSI devices with 1,123,600 logic units and a block random access memory capacity of 62,400 kilobits. It processes infrared prediction data and polarization image streams in parallel via an advanced scalable interface bus. The adaptive optical flow algorithm is embedded as a hardware intellectual property core in the FPGA, employing the Horn-Schuck optical flow calculation principle. Gradient calculation and integration are executed in parallel, completing the processing of the entire 1280 x 1024 pixel area within 140 microseconds. During the displacement compensation calculation, the vibration acceleration integral term is obtained through numerical integration, and the image optical flow term is obtained by calculating the image gradient. The two are fused with a weight of 6:4 to generate the final displacement compensation amount. The vibration acceleration vector is output from step (1) and denoted as . The unit is meters per second squared. For this vector along... , Perform two numerical integrations in the direction ( The effect of directional vibration on planar displacement is negligible. Apply the trapezoidal rule: in The displacement compensation component is obtained by integrating the vibration acceleration. , The calculation was performed through two numerical integrations (using the trapezoidal rule). t0 is the starting time of the integration. t represents the time points during the integration process. , These are the acceleration samples at step k and step (k-1), respectively. Δt k t is the sampling interval, i.e., the time length of the k-th sampling interval; n is the number of integration steps, t k This represents the k-th sampling time. The image optical flow term is obtained by comparing the pixel displacement fields between the 0-degree and 90-degree polarized images. The hardware intellectual property core first calculates the spatial gradients I_x and I_y and the temporal gradient I_t of the two frames, and then solves the following system of linear equations: in , represents the spatial gradient of the image, corresponding to the pixel grayscale gradients in the x and y directions, respectively. The temporal gradient of the image represents the rate of grayscale change between two frames. u and v are the optical flow vector components, corresponding to the optical flow velocities in the x and y directions, respectively. To improve robustness, a multi-scale pyramid structure is embedded within the kernel, iteratively solving at four scale levels, with the smoothing parameter α set to 1.0. Finally, the displacement compensation amount d... comp Generated by fusion of the following formula: Where d opt This is the global average displacement vector extracted from the optical flow field, i.e., the displacement compensation component corresponding to the image optical flow term. This fusion strategy fully utilizes the foresight of vibration prediction and the accuracy of visual correction, effectively overcoming the failure risk of a single mode under strong vibration or extremely low light conditions. The optical flow field fusion calculation stage takes 80 microseconds, meeting the total time constraint of 200 microseconds.

[0036] In the above-mentioned weak light assembly vibration compensation alignment method for electrical connectors, step (4) generates a pulse width modulation control signal based on the displacement compensation amount, driving the piezoelectric ceramic actuator to push the terminal clamping mechanism to move in the opposite direction. The piezoelectric ceramic actuator has a resolution of 0.1 micrometers and a stroke of ±50 micrometers. Specifically, the pulse width modulation control signal is output through a low voltage differential signal interface, the voltage-to-displacement compensation amount conversion coefficient is 0.2 volts per micrometer, and the signal bit width is 12 bits. The piezoelectric ceramic actuator is rigidly integrated into the bottom of the clamping mechanism, and a built-in strain gauge monitors the displacement error in real time. When the error exceeds a set threshold, a fine-tuning cycle is automatically triggered to ensure that the steady-state error is less than 0.3 micrometers. The digital-to-analog conversion module inside the field-programmable gate array converts the 12-bit digital compensation amount into an analog voltage signal with a range of 0 to 4.096 volts, corresponding to a displacement range of 0 to 20.48 micrometers. To cover a stroke of ±50 micrometers, the system uses a bipolar drive circuit to convert the unipolar signal into a -5 to +5 volt output. The piezoelectric ceramic actuator employs a stacked structure, using PZT-5H material with a stiffness of 150 Newtons per micrometer and a resonant frequency greater than 10 kHz. Its response characteristics have been precisely calibrated, with a rise time of 85 microseconds and overshoot less than 5%, ensuring rapid and accurate displacement compensation. The actuator and clamping mechanism are rigidly coupled via Invar connectors, with matched thermal expansion coefficients to prevent additional errors introduced by temperature drift. The compensation command generation stage takes 60 microseconds, including analog-to-digital conversion, drive amplification, and signal transmission.

[0037] In the above-mentioned weak light assembly vibration compensation alignment method for electrical connectors, in step (5), when the absolute value of the displacement compensation is greater than 5 micrometers, a motion correction command is sent to the assembly robot through the Ethernet control automation technology bus, and the motion correction command delay is less than 50 microseconds. Specifically, the Ethernet control automation technology bus transmission protocol ensures that the motion correction command delay is less than 50 microseconds. When the absolute value of the displacement compensation is greater than 5 micrometers, the safety protocol is immediately activated to send the correction command to the assembly robot. The field programmable gate array has a built-in Ethernet media access control layer hard core, which, together with the physical layer chip, achieves 100 megabits per second real-time communication. The motion correction command adopts a simplified message format, which only contains the target coordinate increment (x, y) and timestamp, with a total length of 32 bytes. The protocol stack is implemented at the hardware level, bypassing the operating system interrupt delay. The measured end-to-end delay from compensation determination to completion of physical layer transmission is 42 microseconds. This mechanism is used to handle large-scale disturbances that exceed the stroke range of the piezoelectric ceramic actuator, such as sudden impacts on the conveyor belt or deviations in robot path planning, to ensure the safety and integrity of the system under extreme working conditions.

[0038] In the above-mentioned weak light assembly vibration compensation alignment method for electrical connectors, step (6) uses a built-in strain gauge to form a fine-tuning feedback loop, which provides real-time feedback on displacement error and triggers a fine-tuning cycle to ensure that the compensation accuracy reaches ±0.1 micrometers. Specifically, the fine-tuning feedback loop adopts a closed-loop control strategy, which collects displacement data in real time through the strain gauge, compares it with the preset target value, and generates a correction signal to continuously optimize the compensation accuracy. The strain gauge is attached to the connection between the output end of the piezoelectric ceramic actuator and the clamping mechanism to form a Wheatstone bridge with a sensitivity of 2 millivolts per volt per microstrain. After the signal is amplified by the instrumentation amplifier, it is sampled by a 16-bit analog-to-digital converter with a sampling rate of 10 kHz. The proportional-integral controller in the field-programmable gate array adjusts the error according to the signal. Calculate the correction amount: in For real-time displacement error, The target displacement is the system's preset desired displacement value. The actual displacement data is collected in real time by strain gauges to measure displacement.

[0039] The output correction value of the proportional-integral (PI) controller is used to drive the piezoelectric ceramic actuator for fine displacement adjustment. The proportional gain, with a value of 0.8, is used in the proportional control loop and determines the speed of error response. This is the integral gain, set to 50, used in the integral control loop to eliminate steady-state error. This is the integral term for the error. The error is integrated over time to ensure that the system has no steady-state error.

[0040] When |e(t)| > 0.3 micrometers, the controller initiates a fine-tuning cycle, updating the output every 100 microseconds until the error converges to within ±0.1 micrometers. This feedback loop effectively suppresses the hysteresis nonlinearity of the piezoelectric ceramic and residual external disturbances, ensuring long-term steady-state accuracy.

[0041] To further illustrate the practical application effect of this invention, a specific application scenario is constructed: On an automated assembly line producing micro-board-to-board connectors with a spacing of 0.3 mm, the ambient illuminance is 40 lux, and the conveyor belt motor operates at a frequency of 150 Hz. After the system starts, the infrared thermal imaging sensor detects a temperature rise rate of 2.5 degrees Celsius per second in the motor bearing, determining that the vibration source is activated. The lightweight long short-term memory neural network model outputs a vibration acceleration vector of [0.8, -0.6, 0.2] meters per square second. Simultaneously, the polarized light vision system captures two frames of images, which, after preprocessing, are sent to an adaptive optical flow hardware core to calculate the optical flow displacement as [1.2, -0.9] micrometers. The fused displacement compensation is [1.08, -0.84] micrometers. A field-programmable gate array generates a pulse width modulation signal to drive the piezoelectric ceramic actuator to move in the opposite direction. The strain gauge feedback shows an initial error of 0.25 micrometers, which converges to 0.08 micrometers after three fine-tuning cycles. The entire process takes 198 microseconds, meeting real-time requirements. 30 days of continuous testing showed an average positioning error of 2.7 micrometers, signal crosstalk of -58 dB, and a solder joint failure rate of 0.9%, significantly better than traditional solutions.

[0042] Example 2, based on Example 1, considers an alternative technical solution where vibration prediction does not rely on infrared thermal imaging, but instead uses an embedded accelerometer for direct measurement. (Reference) Figure 2 The dual-modal fusion mechanism framework shown in this embodiment replaces the infrared thermal imaging sensor with a triaxial MEMS accelerometer, model ADXL357, mounted on the surface of the motor housing. This accelerometer has a sampling rate of 4 kHz and a noise density of 25 micrograms per square Hz, communicating with a field-programmable gate array (FPGA) via an SPI bus. The vibration acceleration vector is directly output from the accelerometer, eliminating the neural network prediction stage, but sacrificing foresight—it can only sense vibrations that have already occurred and cannot predict upcoming disturbances. To compensate for this deficiency, the weights of the adaptive optical flow algorithm are adjusted to 30% for the vibration term and 70% for the image optical flow term. Simultaneously, the entire process timing is reassigned: the vibration acquisition phase is shortened to 20 microseconds, polarization imaging remains at 80 microseconds, and the fusion calculation is extended to 100 microseconds, with the total time still controlled within 200 microseconds. This scheme is suitable for scenarios with fixed vibration source locations and stable spectra, and is lower in cost, but its ability to suppress sudden vibrations is weaker than in Embodiment 1. Under the same test conditions, the positioning error was 3.1 micrometers, slightly higher than that of Example 1, but still met the process requirement of ±3 micrometers.

[0043] The above are merely specific embodiments of the present invention, but the technical features of the present invention are not limited thereto. Any simple changes, equivalent substitutions, or modifications made based on the present invention to solve essentially the same technical problems and achieve essentially the same technical effects are all covered within the protection scope of the present invention.

Claims

1. A method for vibration compensation alignment during low-light assembly of electrical connectors, characterized in that: The specific steps include the following: Step 1: Monitor the changes in thermal radiation of the vibration source in real time using an infrared thermal imaging sensor, collect thermal radiation data, and output the vibration acceleration vector through the lightweight long short-term memory neural network model built into the field programmable gate array; Step 2: Synchronously trigger the dual polarization angle imaging of the polarization vision system. The polarization angle switching timing is precisely controlled by the field programmable gate array clock management unit. At 60 microseconds, the 0-degree polarization angle is triggered to capture the reference mark on the shell, and at 65 microseconds, the polarization angle is switched to 90 degrees to enhance the features of the terminal metal surface. Step 3: In the field programmable gate array (FPGA) processing unit, the vibration acceleration vector and polarization image stream are fused together, and the displacement compensation amount is calculated using an adaptive optical flow algorithm; Step 4: Generate a pulse width modulation control signal based on the displacement compensation amount to drive the piezoelectric ceramic actuator to push the terminal clamping mechanism to move in the opposite direction; Step 5: When the absolute value of the displacement compensation is greater than the set value, a motion correction command is sent to the assembly robot via the Ethernet control automation technology bus; Step 6: A fine-tuning feedback loop is formed by using built-in strain gauges to provide real-time feedback on displacement errors and trigger the fine-tuning cycle.

2. The method for weak-light assembly vibration compensation alignment of an electrical connector according to claim 1, characterized in that: The infrared thermal imaging sensor is installed directly above the assembly head at a 45-degree angle to the motor axis. The collected thermal radiation data is an 8-bit grayscale image sequence of 16 x 16 pixels per frame.

3. A method for vibration compensation alignment during weak-light assembly of an electrical connector according to claim 1 or 2, characterized in that: The lightweight long short-term memory neural network model has fewer than 10,000 parameters, takes 10 consecutive frames of thermal radiation image sequence as input, and outputs a three-dimensional vibration acceleration vector.

4. The method for weak-light assembly vibration compensation alignment of an electrical connector according to claim 1, characterized in that: The polarized light vision system adopts a ring layout and includes eight groups of 850-nanometer light-emitting diodes. The polarization angle can be switched between 0 degrees and 90 degrees. It is coaxially mounted with the image sensor and is 20 centimeters away from the terminal contact surface.

5. The method for vibration compensation alignment of electrical connectors in low-light assembly according to claim 4, characterized in that: During the dual polarization angle imaging process, the time difference between the acquisition of two frames is less than 5 microseconds. The 0-degree polarization angle imaging is used to suppress ambient light interference, and the 90-degree polarization angle imaging enhances the reflection characteristics of the terminal metal surface through Brewster angle.

6. The method for vibration compensation alignment of electrical connectors in low-light assembly according to claim 1, characterized in that: The field-programmable gate array uses Xilinx VLSI devices and processes infrared prediction data and polarization image streams in parallel through an advanced scalable interface stream bus.

7. The method for vibration compensation alignment of electrical connectors in low-light assembly according to claim 1, characterized in that: The adaptive optical flow algorithm is embedded as a hardware intellectual property core in a field-programmable gate array. It adopts the Horn-Schock optical flow calculation principle, and the gradient calculation and integral operation are executed in parallel. It completes the processing of the entire area of ​​1280 by 1024 pixels in 140 microseconds.

8. The method for vibration compensation alignment of electrical connectors in low-light assembly according to claim 1, characterized in that: The pulse width modulation control signal is output through a low-voltage differential signal interface, with a voltage-to-displacement compensation conversion coefficient of 0.2 volts per micrometer and a signal bit width of 12 bits.

9. The method for vibration compensation alignment of electrical connectors in low-light assembly according to claim 1, characterized in that: The piezoelectric ceramic actuator is rigidly integrated into the bottom of the clamping mechanism, with built-in strain gauges to monitor displacement error in real time. When the error exceeds the set threshold, a fine-tuning cycle is automatically triggered to ensure that the steady-state error is less than 0.3 micrometers.

10. The method for vibration compensation alignment of electrical connectors in low-light assembly according to claim 1, characterized in that: The method operates under low-light conditions with an illumination level below 50 lux, with the entire process time strictly controlled within 200 microseconds, and supports a full range of connectors with terminal pitches from 0.2 to 1.0 mm.

Images