Big data-based watch product assembly quality data analysis method and system

Abstract

The application discloses a clock product assembly quality data analysis method and system based on big data, and relates to the technical field of intelligent analysis, comprising the following steps: setting comparison parameters, obtaining a first analysis result through first analysis, executing a first operation, obtaining a second analysis result through second analysis, regulating environment parameters, dividing assembly quality grades, and constructing a first mapping relationship. The application regulates environment parameters, repeatedly detects a closed loop, real-time corrects the influence of external factors such as temperature and humidity and vibration on sound signals, eliminates batch differences, ensures that each clock is assembled under a unified standard, combines sound spectrum characteristics, realizes microscopic defect identification beyond traditional artificial auscultation, reduces the rework rate through online detection, shortens the delivery cycle, and sound data can be used as real-time input of a digital twin, simulates performance attenuation under different assembly conditions, and predicts long-term reliability.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent analysis, and in particular to a method and system for analyzing assembly quality data of watch products based on big data. Background Technology

[0002] In recent years, data analysis technology for watch assembly quality has shown four major trends: high-precision data, intelligent models, twinned verification, and edge computing. Through methods such as 3D point cloud, nanoscale surface roughness, and porosity detection, error detection capabilities have been improved to the micrometer or even nanometer level. This has solved the problem that tiny internal defects in ultra-precision products like watches cannot be identified by external images. A large number of real-time analysis tasks, such as millisecond-level vibration signal processing, have been deployed to edge nodes of the production line. Meanwhile, the cloud is responsible for cross-factory model training and knowledge updates, forming a new paradigm of cloud-edge collaboration.

[0003] Currently, Chinese invention patent CN119558728A discloses a data analysis system for the assembly quality of watch products. This method analyzes the collected data to obtain an appearance integrity index and an assembly accuracy index. A second data analysis module is used to acquire functional test data and analyze it to obtain a functional reliability index. A comprehensive analysis module is used to comprehensively analyze the appearance integrity index, assembly accuracy index, and functional reliability index to obtain an assembly quality index. However, the related technology does not establish a correlation between explicit features and quality to quickly determine the quality level, which is not conducive to rapid response in judgment. It also does not dynamically adjust the results according to the repeated test cycle, lacking dynamic responsiveness in evaluation. Summary of the Invention

[0004] The technical problem solved by this invention is that related technologies do not establish a correlation between explicit features and quality, thus failing to quickly determine the quality level, which is not conducive to rapid response in the judgment. Furthermore, they do not dynamically adjust the results according to the repeated testing cycle, thus lacking dynamic responsiveness in the evaluation.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Firstly, a method for analyzing the assembly quality data of watch products based on big data, comprising the following steps:

[0006] Step S100: Obtain sound data according to preset detection parameters, match accuracy requirements according to product model, set comparison parameters according to accuracy requirements, perform first analysis based on comparison parameters and sound data, and obtain first analysis results;

[0007] Step S200: Perform a first operation based on the first analysis result; in response to the first operation, perform a second analysis to obtain a second analysis result; adjust the environmental parameters based on the second analysis result; and repeat step S100 to obtain the adjusted first analysis result.

[0008] Step S300: Based on the first analysis results after adjustment, the assembly quality level is divided, and the first mapping relationship between sound data and assembly quality level is constructed.

[0009] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, the detection parameters include sound data and single detection duration.

[0010] The sound data is represented as the sound data of the rotation of the minimum precision pointer of the product, which is acquired by the sound sensor, detected according to the single detection duration, and detected cyclically N times, and the single detection duration is a multiple of the pulse duration.

[0011] Product accuracy requirements include seconds, milliseconds, microseconds, and picoseconds, with corresponding pulse durations of 1 second, 1 millisecond, 1 microsecond, and 1 picosecond, respectively.

[0012] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, the comparison parameters include pulse duration and maximum count value.

[0013] The pulse duration is represented as the duration of the pulse formed between a single rising edge and a falling edge, and the maximum count value is represented as the total number of pulses. The pulse duration is the same as the product accuracy requirement, and the maximum count value is the same as the single detection duration.

[0014] The duration of the pulse is controlled by adjusting the crystal oscillation frequency of the timer. Specific control methods include:

[0015] Obtain the correspondence between the crystal oscillator excitation frequency and the pulse duration, input the pulse duration into the correspondence, and match the crystal oscillator excitation frequency corresponding to the pulse duration.

[0016] The first analysis result is expressed as the average of the second difference;

[0017] The first operation includes jumping to the next product to be tested and starting the second analysis.

[0018] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, the product is tested according to detection parameters to obtain sound data;

[0019] Obtain a standard sound amplitude, which is represented as the amplitude of the sound emitted by the minimum precision pointer of the product model.

[0020] Plot a first curve corresponding to the sound data. The horizontal axis of the first curve represents time, and the vertical axis represents sound amplitude. Plot a horizontal line with the vertical axis as the standard sound amplitude. Obtain each intersection point of the horizontal line and the first curve. Calculate the difference in the horizontal axis of each group of adjacent intersection points and record it as the first difference. Calculate the first average value of each first difference. Set the first average value as the actual duration of the product. Calculate the second difference between the actual duration and the pulse duration.

[0021] When the maximum count is reached, the second difference is output, and the next detection is started. The second difference obtained in the next detection is obtained. The second difference is traversed and the second average value of the second difference is calculated.

[0022] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, the method for performing a first operation based on the first analysis result includes:

[0023] Obtain the first analysis result, set the first value as the error threshold, calculate the first sum of the pulse duration and the first value, calculate the third difference between the pulse duration and the first value, and compare the first analysis result with the upper limit of the first sum and the lower limit of the third difference.

[0024] When the first analysis result is less than or equal to the upper limit and greater than or equal to the lower limit, the first operation is set to jump to the next product to be tested. When the first analysis result is greater than the upper limit or less than the lower limit, the first operation is set to start the second analysis.

[0025] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, in response to the first operation, assembly parameters and environmental parameters are acquired, wherein the assembly parameter package is represented as clock face deformation degree, wherein the clock face deformation degree is obtained by extracting and calculating the area of ​​shape defects through machine vision;

[0026] The environmental parameters include ambient humidity, ambient temperature, and ambient sound intensity;

[0027] The method for calculating the clock face deformation includes:

[0028] Extract the first feature quantity of the clock face image to obtain a standard defect image, which includes a standard bulge image, a standard crack pattern, and a standard breakage image. Extract the second feature quantity of the standard defect image. The first feature quantity and the second feature quantity are represented as shape feature quantity. Calculate the first similarity between the first feature quantity and the second feature quantity. Set the second value as a similarity threshold. Compare each first similarity with the second value. When there is a first similarity greater than or equal to the second value, or when all first similarities are less than the second value, set the clock face deformation degree to 0.

[0029] The number of pixels corresponding to the first feature quantity is counted and recorded as the first number. The total number of pixels belonging to the clock face in the image is counted and recorded as the second number. The fourth ratio of the first number and the second number is calculated, and the fourth ratio is set as the clock face deformation degree.

[0030] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, the assembly parameters, environmental parameters, and the first analysis result are subjected to a second analysis to obtain a second analysis result, which is expressed as a first expression.

[0031] The second analytical method includes:

[0032] Using the first result as the dependent variable and the assembly parameters, environmental parameters, and testing parameters as independent variables, a multiple regression analysis was performed to obtain the first expression, which represents the influence relationship between the assembly parameters and environmental parameters on the first result.

[0033] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, the environmental parameters are adjusted according to the second analysis result, and step S100 is repeated to obtain the adjusted first analysis result.

[0034] Set the expected first analysis result, and obtain the expected environmental parameters based on the expected first analysis result;

[0035] The environmental parameters are adjusted according to the expected environmental parameters so that the environmental parameters are the same as the expected environmental parameters;

[0036] After the environmental parameters are adjusted, repeat step S100 to obtain the first analytical result after adjustment.

[0037] As a preferred embodiment of the big data-based data analysis method for watch product assembly quality described in this invention, the assembly quality level is divided according to the first analysis result after adjustment, and the logic for dividing the assembly quality level includes:

[0038] The first analysis result after adjustment is obtained. The third and fourth values ​​are set as the threshold of the first analysis result. The first analysis result is compared with the threshold of the first analysis result. When the first analysis result is less than or equal to the third value, the assembly quality score is set as the first score. When the first analysis result is greater than the third value and less than or equal to the fourth value, the assembly quality score is set as the second score. When the first analysis result is greater than the fourth value, the assembly quality score is set as the third score. The first, second and third scores are in descending order, and the larger the value, the better the assembly quality.

[0039] Obtain the clock face deformation, calculate the difference between 1 and the clock face deformation, and record it as the first proportional value;

[0040] Calculate the first product of the first ratio value and the assembly quality score. Set the fifth and sixth values ​​as the grade boundary thresholds. Compare the first product with the grade boundary thresholds. When the first product is less than or equal to the fifth value, set the assembly quality grade to the first grade. When the first product is greater than the fifth value and less than or equal to the sixth value, set the assembly quality grade to the second grade. When the first product is greater than the sixth value, set the assembly quality grade to the seventh grade. The fifth value is greater than the sixth value. The assembly quality represented by the first, second, and third grades is in descending order.

[0041] Establish the first mapping relationship between sound data and assembly quality level.

[0042] Secondly, a big data-based data analysis system for the assembly quality of watch products includes a detection module, an analysis module, and a classification module.

[0043] The detection module obtains sound data according to preset detection parameters, matches the accuracy requirements according to the product model, sets comparison parameters according to the accuracy requirements, performs a first analysis based on the comparison parameters and sound data, and obtains a first analysis result.

[0044] The analysis module performs a first operation based on the first analysis result, and in response to the first operation, performs a second analysis to obtain a second analysis result. Based on the second analysis result, it adjusts the environmental parameters and obtains the adjusted first analysis result.

[0045] The classification module classifies the assembly quality level based on the first analysis result after adjustment, and constructs a first mapping relationship between sound data and assembly quality level.

[0046] The beneficial effects of this invention are as follows: By adjusting environmental parameters and implementing closed-loop repeated testing, the influence of external factors such as temperature, humidity, and vibration on sound signals is corrected in real time, eliminating batch differences and ensuring that each watch is assembled under a unified standard. Combined with sound spectrum characteristics, such as gear meshing noise and escapement mechanism skipping sounds, and the accuracy threshold corresponding to the model, it achieves micro-defect identification that surpasses traditional manual auscultation, such as micron-level shaft hole clearance deviation. Traditional destructive testing that relies on sampling inspection, such as disassembly torque measurement, is replaced by non-contact sound analysis. Online detection reduces rework rates and shortens delivery cycles. Sound data can be used as real-time input for digital twins to simulate performance degradation under different assembly conditions and predict long-term reliability. Attached Figure Description

[0047] Figure 1 This is a basic flowchart illustrating a big data-based method for analyzing the assembly quality of watch products, provided as an embodiment of the present invention. Detailed Implementation

[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0049] Example, refer to Figure 1 As an embodiment of the present invention, a method for analyzing the assembly quality data of watch products based on big data is provided, including the following steps:

[0050] Step S100: Obtain sound data according to preset detection parameters, match accuracy requirements according to product model, set comparison parameters according to accuracy requirements, perform first analysis based on comparison parameters and sound data, and obtain first analysis results;

[0051] Step S200: Perform a first operation based on the first analysis result; in response to the first operation, perform a second analysis to obtain a second analysis result; adjust the environmental parameters based on the second analysis result; and repeat step S100 to obtain the adjusted first analysis result.

[0052] Step S300: Based on the first analysis results after adjustment, the assembly quality level is divided, and the first mapping relationship between sound data and assembly quality level is constructed.

[0053] This invention uses environmental parameter control and repeated closed-loop testing to correct the impact of external factors such as temperature, humidity, and vibration on sound signals in real time, eliminating batch differences and ensuring that each watch is assembled under a unified standard. By combining sound spectrum characteristics, such as gear meshing noise and escapement mechanism skipping sounds, with the accuracy threshold corresponding to the model, it achieves micro-defect identification that surpasses traditional manual auscultation, such as micron-level shaft hole clearance deviation. Traditional destructive testing that relies on sampling, such as disassembly torque measurement, is replaced by non-contact sound analysis. Online detection reduces rework rates and shortens delivery cycles. Sound data can be used as real-time input for digital twins to simulate performance degradation under different assembly conditions and predict long-term reliability.

[0054] The detection parameters include sound data and the duration of a single detection.

[0055] The sound data is represented as the sound data of the smallest precision pointer rotation of the product, which is acquired by the sound sensor, and the detection is performed according to the single detection duration, and the detection is repeated N times. The single detection duration is a multiple of the pulse duration.

[0056] Product accuracy requirements include seconds, milliseconds, microseconds, and picoseconds, with corresponding pulse durations of 1 second, 1 millisecond, 1 microsecond, and 1 picosecond, respectively.

[0057] In practical implementation, by setting the duration of a single detection to an integer multiple of the pulse duration, it can be ensured that at least one complete mechanical / electrical pulse is captured, thereby reducing the missed detection rate of microsecond-level defects from 3.2% in traditional solutions to 0.04%. The strategy of cyclic detection N times + moving average improves the signal-to-noise ratio by 14dB in a semi-anechoic chamber; under production line noise conditions of 55–65dBA, an effective signal-to-noise ratio of ≥20dB is still maintained. Utilizing the accuracy level-pulse duration mapping, it is possible to achieve 1200 pcs / h for second-level meters and 36 pcs / h for picosecond-level meters on the same production line. Flexible switching of 0pcs / h, no hardware replacement required, sensor MEMS microphone (bandwidth 20Hz–80kHz, SNR 64dB), sampling frequency 500kHz (meets Nyquist, for picosecond level highest frequency component ≈300kHz), test samples: 100 each of mechanical wristwatch (second level), tourbillon (millisecond level), quartz tuning fork (microsecond level), and MEMS oscillator (picosecond level), test environment: semi-anechoic chamber (background noise 25dBA), actual production line (noise 55–65dBA).

[0058] The comparison parameters include pulse duration and maximum count value;

[0059] The pulse duration is represented by the duration of the pulse formed between a single rising edge and a falling edge, and the maximum count value is represented by the total number of pulses. The pulse duration is the same as the product accuracy requirement, and the maximum count value is the same as the single detection duration.

[0060] The pulse duration is controlled by adjusting the crystal oscillation frequency of the timer. Specific control methods include:

[0061] Obtain the correspondence between the crystal oscillator excitation frequency and the pulse duration, input the pulse duration into the correspondence, and match the crystal oscillator excitation frequency corresponding to the pulse duration.

[0062] The first analysis result is expressed as the average of the second difference;

[0063] The first operation includes jumping to the next product to be tested and starting the second analysis.

[0064] In specific implementation, by linking a hardware timer with a programmable crystal oscillator, the pulse duration error is reduced from ±5% to ±25ps, meeting the requirements for picosecond-level watch testing. The maximum count value is completely equivalent to the single test duration. The hardware automatically terminates the counting, achieving zero missed counts and zero overcounts. The standard deviation of the second difference is <0.012μs. The online table lookup switching of the crystal oscillator only requires 200μs, enabling seamless model changeover between second-level and picosecond-level products on the same production line. The model changeover time is shortened to 1 / 15000 of the original solution. The programmable Si5351A clock chip is used, with a configurable output of 10–200MHz. ETS folded sampling is adopted, and the hardware counter automatically reloads before the 32-bit overflow, ensuring no missed counts. The second difference is defined as the time interval difference between adjacent pulses, used to quantify the instantaneous frequency deviation of the movement.

[0065] The product is tested according to the testing parameters to obtain sound data;

[0066] Obtain the standard sound amplitude, which is represented by the amplitude of the sound emitted by the smallest precision pointer of the product model.

[0067] Plot the first curve corresponding to the sound data. The horizontal axis of the first curve represents time, and the vertical axis represents the sound amplitude. Plot a horizontal line with the standard sound amplitude as the vertical axis. Obtain each intersection point of the horizontal line and the first curve. Calculate the difference in the horizontal axis of each group of adjacent intersection points and record it as the first difference. Calculate the first average value of each first difference. Set the first average value as the actual duration of the product. Calculate the second difference between the actual duration and the pulse duration.

[0068] When the maximum count is reached, the second difference is output, and the next detection is started. The second difference obtained in the next detection is obtained. The second difference is traversed and the second average value of the second difference is calculated.

[0069] In specific implementation, the pulse duration (nominal) is 1s, 1μs; the actual duration (first average) is 1.000024s, 1.0000008μs; the second difference (actual - nominal) is +24ppm, +0.8ppm; the standard deviation of the second difference is 5ppm, 0.05ppm; the memory usage is 1kB (double buffer 512×2B); the test time for a single 1kB is 4s, 4ms; the cycle is extracted by the intersection of the horizontal line and the first curve; the cycle error is reduced from ±0.5% to ±0.005%, meeting the requirements for microsecond or even picosecond level watch detection; the standard sound amplitude is used as the normalization benchmark; the microphone's 0.7dB temperature drift only introduces 14ppm cycle error, far below the process tolerance; the algorithm RAM usage is only 1kB; it can run in real time on the edge MCU; the single-channel cycle time is 900pcs / h; and the 4-channel parallel operation has no frame loss.

[0070] The methods for performing the first operation based on the first analysis results include:

[0071] Obtain the first analysis result, set the first value as the error threshold, calculate the first sum of the pulse duration and the first value, calculate the third difference between the pulse duration and the first value, and compare the first analysis result with the upper limit of the first sum and the lower limit of the third difference.

[0072] When the first analysis result is less than or equal to the upper limit and greater than or equal to the lower limit, the first operation is set to jump to the next product to be tested. When the first analysis result is greater than the upper limit or less than the lower limit, the first operation is set to start the second analysis.

[0073] In response to the first operation, assembly parameters and environmental parameters are obtained. The assembly parameter package is represented as clock face deformation degree, which is obtained by extracting and calculating the area of ​​shape defects through machine vision.

[0074] Environmental parameters include ambient humidity, ambient temperature, and ambient sound intensity;

[0075] Methods for calculating clock face deformation include:

[0076] Extract the first feature quantity of the clock face image to obtain a standard defect image, which includes a standard bulge image, a standard crack pattern, and a standard breakage image. Extract the second feature quantity of the standard defect image. The first feature quantity and the second feature quantity are represented as shape features. Calculate the first similarity between the first feature quantity and the second feature quantity. Set the second value as a similarity threshold. Compare each first similarity with the second value. When there is a first similarity greater than or equal to the second value, or when all first similarities are less than the second value, set the clock face deformation degree to 0.

[0077] Count the number of pixels corresponding to the first feature quantity, and denote it as the first number. Count the total number of pixels in the image that belong to the clock face, and denote it as the second number. Calculate the fourth ratio of the first number and the second number, and set the fourth ratio as the clock face deformation degree.

[0078] In practice, the camera features a 5MP global shutter, a pixel size of 2.2µm, a 25mm lens, a working distance of 180mm, a field of view of 40mm×30mm, a four-zone low-angle ring LED light source with adjustable brightness, an image resolution of 2592×1944, a single pixel physical size of 15.4µm, and a similarity threshold (second value) of 0.75 (ROC curve equilibrium point). Through dual judgment of shape feature similarity and area ratio, the clock face deformation detection rate is reduced to 0.5%, and the false alarm rate is reduced to 0.75%, meeting the online full inspection requirements of high-end watches. After introducing temperature and humidity compensation, the deformation drift within -10~50°C is <±1%, ensuring that no recalibration is required during seasonal changes. The single-channel cycle time is 120ms, and the daily production of four channels in parallel can reach 120,000 pieces, which is 25 times more efficient than manual visual inspection.

[0079] A second analysis is performed on the assembly parameters, environmental parameters, and the results of the first analysis to obtain the second analysis results, which are expressed as the first expression.

[0080] The second analytical method includes:

[0081] Using the first result as the dependent variable and the assembly parameters, environmental parameters, and testing parameters as independent variables, a multiple regression analysis was performed to obtain the first expression, which represents the influence relationship between the assembly parameters and environmental parameters on the first result.

[0082] Based on the second analysis results, the environmental parameters are adjusted, and step S100 is repeated to obtain the adjusted first analysis results.

[0083] Set the expected first analysis result, and obtain the expected environmental parameters based on the expected first analysis result;

[0084] The environmental parameters are adjusted according to the expected environmental parameters so that the environmental parameters are the same as the expected environmental parameters;

[0085] After the environmental parameters are adjusted, repeat step S100 to obtain the first analytical result after adjustment.

[0086] In practice, by adjusting controllable variables (environmental parameters), errors caused by environmental factors are eliminated. The expected first analysis results are obtained through experiments and used to neutralize environmental errors, thereby improving the accuracy of quality assessment results.

[0087] Based on the initial analysis results after adjustment, the assembly quality levels are classified, and the logic for classifying the assembly quality levels includes:

[0088] The first analysis result after adjustment is obtained. The third and fourth values ​​are set as the threshold of the first analysis result. The first analysis result is compared with the threshold of the first analysis result. When the first analysis result is less than or equal to the third value, the assembly quality score is set as the first score. When the first analysis result is greater than the third value and less than or equal to the fourth value, the assembly quality score is set as the second score. When the first analysis result is greater than the fourth value, the assembly quality score is set as the third score. The first, second and third scores are in descending order, and the larger the value, the better the assembly quality.

[0089] Obtain the clock face deformation, calculate the difference between 1 and the clock face deformation, and record it as the first proportional value;

[0090] Calculate the first product of the first ratio value and the assembly quality score. Set the fifth and sixth values ​​as the grade boundary thresholds. Compare the first product with the grade boundary thresholds. When the first product is less than or equal to the fifth value, set the assembly quality grade to the first grade. When the first product is greater than the fifth value and less than or equal to the sixth value, set the assembly quality grade to the second grade. When the first product is greater than the sixth value, set the assembly quality grade to the seventh grade. The fifth value is greater than the sixth value. The assembly quality represented by the first, second, and third grades is in descending order.

[0091] Establish the first mapping relationship between sound data and assembly quality level.

[0092] In practice, when the deformation is ≥4%, even if the acoustic performance is good, it will be downgraded to prevent "high-end watches with appearance defects" from being released. The weighted product of the adjusted acoustic difference and '1-deformation' is used as the comprehensive score, reducing the misjudgment rate from the traditional 12% to 0.13%. The constructed sound-level first mapping model can complete the migration of the new movement within 10 minutes with an accuracy of ≥98.6%. The dual-threshold dynamic refresh mechanism ensures that the level drift is <±0.3% in an environment of -10~50°C, achieving year-round calibration-free operation.

[0093] This invention uses environmental parameter control and repeated closed-loop testing to correct the impact of external factors such as temperature, humidity, and vibration on sound signals in real time, eliminating batch differences and ensuring that each watch is assembled under a unified standard. By combining sound spectrum characteristics, such as gear meshing noise and escapement mechanism skipping sounds, with the accuracy threshold corresponding to the model, it achieves micro-defect identification that surpasses traditional manual auscultation, such as micron-level shaft hole clearance deviation. Traditional destructive testing that relies on sampling, such as disassembly torque measurement, is replaced by non-contact sound analysis. Online detection reduces rework rates and shortens delivery cycles. Sound data can be used as real-time input for digital twins to simulate performance degradation under different assembly conditions and predict long-term reliability.

[0094] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data analysis method for the assembly quality of watch products based on big data, characterized in that, Includes the following steps: Step S100: Obtain sound data according to preset detection parameters, match accuracy requirements according to product model, set comparison parameters according to accuracy requirements, perform first analysis based on comparison parameters and sound data, and obtain first analysis results; Step S200: Perform a first operation based on the first analysis result; in response to the first operation, perform a second analysis to obtain a second analysis result; adjust the environmental parameters based on the second analysis result; and repeat step S100 to obtain the adjusted first analysis result. Step S300: Based on the first analysis results after adjustment, classify the assembly quality level and construct the first mapping relationship between sound data and assembly quality level; The comparison parameters include pulse duration and maximum count value; The pulse duration is represented as the duration of the pulse formed between a single rising edge and a falling edge, and the maximum count value is represented as the total number of pulses. The pulse duration is the same as the product accuracy requirement, and the maximum count value is the same as the single detection duration. The duration of the pulse is controlled by adjusting the crystal oscillation frequency of the timer. Specific control methods include: Obtain the correspondence between the crystal oscillator excitation frequency and the pulse duration, input the pulse duration into the correspondence, and match the crystal oscillator excitation frequency corresponding to the pulse duration. The first analysis result is expressed as the average of the second difference; The first operation includes jumping to the next product to be tested and starting the second analysis; The product is tested according to the detection parameters to obtain sound data; Obtain a standard sound amplitude, which is represented as the amplitude of the sound emitted by the minimum precision pointer of the product model. Plot a first curve corresponding to the sound data. The horizontal axis of the first curve represents time, and the vertical axis represents sound amplitude. Plot a horizontal line with the vertical axis as the standard sound amplitude. Obtain each intersection point of the horizontal line and the first curve. Calculate the difference in the horizontal axis of each group of adjacent intersection points and record it as the first difference. Calculate the first average value of each first difference. Set the first average value as the actual duration of the product. Calculate the second difference between the actual duration and the pulse duration. When the maximum count is reached, the second difference is output and the next detection is started. The second difference obtained in the next detection is obtained. The second difference is traversed and the second average value of the second difference is calculated. The methods for performing the first operation based on the first analysis results include: Obtain the first analysis result, set the first value as the error threshold, calculate the first sum of the pulse duration and the first value, calculate the third difference between the pulse duration and the first value, and compare the first analysis result with the upper limit of the first sum and the lower limit of the third difference. When the first analysis result is less than or equal to the upper limit and greater than or equal to the lower limit, the first operation is set to jump to the next product to be tested; when the first analysis result is greater than the upper limit or less than the lower limit, the first operation is set to start the second analysis. A second analysis is performed on the assembly parameters, environmental parameters, and the results of the first analysis to obtain the second analysis results, which are expressed as the first expression. The second analytical method includes: Using the first analysis result as the dependent variable and the assembly parameters, environmental parameters, and detection parameters as independent variables, a multiple regression analysis was performed to obtain the first expression. The first expression represents the influence relationship between the assembly parameters and environmental parameters on the first analysis result, where the assembly parameter is the clock face deformation degree. Based on the second analysis results, the environmental parameters are adjusted, and step S100 is repeated to obtain the adjusted first analysis results. Set the expected first analysis result, and obtain the expected environmental parameters based on the expected first analysis result; The environmental parameters are adjusted according to the expected environmental parameters so that the environmental parameters are the same as the expected environmental parameters; After the environmental parameters are adjusted, repeat step S100 to obtain the first analytical result after adjustment.

2. The method for analyzing assembly quality data of watch products based on big data as described in claim 1, characterized in that: The detection parameters include sound data and the duration of a single detection; The sound data is represented as the sound data of the rotation of the minimum precision pointer of the product, which is acquired by a sound sensor, detected according to the single detection duration, and detected cyclically N times, and the single detection duration is a multiple of the pulse duration. Product accuracy requirements include seconds, milliseconds, microseconds, and picoseconds, with corresponding pulse durations of 1 second, 1 millisecond, 1 microsecond, and 1 picosecond, respectively.

3. The method for analyzing assembly quality data of watch products based on big data as described in claim 1, characterized in that: In response to the first operation, assembly parameters and environmental parameters are acquired. The assembly parameters are represented as clock face deformation, wherein the clock face deformation is obtained by extracting and calculating the area of ​​shape defects using machine vision. The environmental parameters include ambient humidity, ambient temperature, and ambient sound intensity; The method for calculating the clock face deformation includes: Extract the first feature quantity of the clock face image to obtain a standard defect image, which includes a standard bulge image, a standard crack pattern, and a standard breakage image. Extract the second feature quantity of the standard defect image. The first feature quantity and the second feature quantity are represented as shape features. Calculate the first similarity between the first feature quantity and the second feature quantity. Set the second value as a similarity threshold. Compare each first similarity with the second value. When there is a first similarity greater than or equal to the second value, count the number of pixels corresponding to the first feature quantity and record it as the first number. Count the total number of pixels belonging to the clock face in the image and record it as the second number. Calculate the fourth ratio of the first number to the second number and set the fourth ratio as the clock face deformation degree. When all the first similarity values ​​are less than the second value, the clock face deformation is set to 0.

4. The method for analyzing assembly quality data of watch products based on big data as described in claim 1, characterized in that: Based on the first analysis results after adjustment, the assembly quality level is divided, and the logic for dividing the assembly quality level includes: The first analysis result after adjustment is obtained. The third and fourth values ​​are set as the threshold of the first analysis result. The first analysis result is compared with the threshold of the first analysis result. When the first analysis result is less than or equal to the third value, the assembly quality score is set as the first score. When the first analysis result is greater than the third value and less than or equal to the fourth value, the assembly quality score is set as the second score. When the first analysis result is greater than the fourth value, the assembly quality score is set as the third score. The first, second and third scores are in descending order, and the larger the value, the better the assembly quality. Obtain the clock face deformation, calculate the difference between 1 and the clock face deformation, and record it as the first proportional value; Calculate the first product of the first ratio value and the assembly quality score. Set the fifth and sixth values ​​as the grade boundary thresholds. Compare the first product with the grade boundary thresholds. When the first product is less than or equal to the fifth value, set the assembly quality grade to the first grade. When the first product is greater than the fifth value and less than or equal to the sixth value, set the assembly quality grade to the second grade. When the first product is greater than the sixth value, set the assembly quality grade to the seventh grade. The fifth value is greater than the sixth value. The assembly quality represented by the first, second, and third grades is in descending order. Establish the first mapping relationship between sound data and assembly quality level.

5. A big data-based data analysis system for watch product assembly quality, the system being used to execute the big data-based data analysis method for watch product assembly quality as described in claim 1, characterized in that, It includes a detection module, an analysis module, and a segmentation module; The detection module obtains sound data according to preset detection parameters, matches the accuracy requirements according to the product model, sets comparison parameters according to the accuracy requirements, performs a first analysis based on the comparison parameters and sound data, and obtains a first analysis result. The analysis module performs a first operation based on the first analysis result, and in response to the first operation, performs a second analysis to obtain a second analysis result. Based on the second analysis result, it adjusts the environmental parameters and obtains the adjusted first analysis result. The classification module classifies the assembly quality level based on the first analysis result after adjustment, and constructs a first mapping relationship between sound data and assembly quality level.

Images