Mobile phone rotating shaft torsion force matching method

Abstract

This invention discloses a method for matching the torque of a mobile phone hinge, comprising the following steps: constructing and training an AI torque matching model, digitizing and binding data for traceability, formulating AI pairing rules, automating the selection and placement of multiple material trays using AI priority, automating execution and integrating with the production line, and dynamically optimizing and adjusting in real time. The method involves collecting torque mechanical data of small components of the mobile phone hinge, preprocessing the data to divide it into mechanical intervals, statistically analyzing the data volume in each interval and performing a normality test to determine valid and invalid data; constructing an AI torque component matching model based on machine learning algorithms, using the torque characteristics of the small components as input and the predicted torque of the entire hinge as output, training and optimizing the model parameters to improve the model's prediction accuracy and robustness. This invention, through precise AI matching, controls the summation torque tolerance of the entire hinge after matching the small hinge components to within 20N, solving the problems of large torque fluctuations and wide torque range of the entire hinge in the prior art, and significantly improving the consistency of the feel of the mobile phone hinge.

Description

Technical Field

[0001] This invention relates to the field of mobile phone component processing and assembly technology, and in particular to a method for matching the torque of a mobile phone hinge. Background Technology

[0002] The feel of a phone hinge is a key factor affecting user experience, and its torque consistency directly determines users' evaluation and choice of the product. In the current production process of small phone hinge components, the torque fluctuation range is 115-165N, resulting in an overall hinge torque range of 230-330N and a torque tolerance range of 100N. There is significant room for improvement in the consistency of the feel.

[0003] Meanwhile, the existing assembly lines for small shaft components use a combination of automated lines and manual pairing, which has many technical defects: the small components are not laser-engraved with QR codes, the material trays and blister boxes lack error-proofing measures, the data of the small components is not bound to the physical components in a 1:1 ratio, the material trays are not effectively distinguished, and intelligent pairing calculation and material quantity ratio sorting are not introduced, resulting in low efficiency and poor matching accuracy of manual pairing, and the production yield and automation level are limited due to component mismatch.

[0004] Traditional torque matching methods lack intelligent data analysis and matching means, making it impossible to achieve precise matching based on the torque characteristics of small components. Furthermore, they are difficult to adapt to dynamic changes in data during the production process, resulting in poor replicability of matching solutions and making it impossible to extend them to other component matching scenarios. This hinders the digital and intelligent upgrading of factories.

[0005] Therefore, how to provide a method for matching the torque of a mobile phone hinge is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One object of the present invention is to provide a method for matching the torque of a mobile phone hinge, comprising the following steps: Torque data acquisition and preprocessing Torque mechanical data of a mobile phone hinge component was collected. After collection, the data was preprocessed, divided into different torque mechanical intervals, and the data volume of each interval was counted. The overall data was subjected to the Anderson-Darling normality test, and the characteristics of the data such as mean, median, standard deviation, skewness, and kurtosis were analyzed. Invalid data was removed. The invalid data were the low interval data where the sum of the pairwise torque values ​​could not fall into the set full shaft target torque range. The amount of valid data was determined to provide a data foundation for subsequent AI pairing.

[0007] AI Torque Selection Model Construction and Training Based on machine learning algorithms, including neural networks and support vector machines, an AI torque component selection model is constructed. The model is trained by taking the collected torque characteristic data of small components as input and the predicted value of the whole shaft torque after different component combinations as output. The model parameters are continuously optimized to improve the prediction accuracy and stability of the model. At the same time, various interference factors in the production process are incorporated to enhance the robustness of the algorithm and make the model adaptable to actual production scenarios.

[0008] Data digitization binding and traceability The process of assembling small mobile phone hinge components has been optimized. In the processes of component testing and feeding, torque and angle testing, and sorting and traying, QR codes are laser-engraved on the components, and error-proofing measures are added to the trays and blister boxes. This achieves a 1:1 binding between component data and physical components, and uploads the torque test results of the components to the factory's MES system in real time. The torque data can be traced and retrieved throughout the entire process through QR codes.

[0009] AI matching rule setting The trained AI model is used to retrieve historical torque data from the MES system and calculate the median of the data distribution. In combination with product design requirements, the target torque range of the whole shaft is set, and pairing rules for small components are formulated. That is, when the sum of the torque values ​​of two small components falls within the set range, it is the optimal pairing. At the same time, a relaxed range is set for pairing and utilization of the remaining data.

[0010] AI-driven automated selection and placement of multiple trays with priority settings Configure multiple pairing trays and 1 defective tray. All trays are set to an even-numbered column structure. Pairing is completed when the torque values ​​of adjacent small components in two columns meet the mechanical specifications set by AI, and the pairing columns are matched with each other from left to right. The multi-tray priority pairing mechanism is implemented: with the pairing tray 1 as the first priority, the AI ​​model first fills the pairing column of tray 1 with the small components that meet the optimal pairing conditions. After the waiting queue of tray 1 is full, the small components that do not meet the optimal pairing conditions are transferred to the waiting queue of tray 2 and tray 3 in turn. After tray 1 is full and the tray is closed, tray 2 and tray 3 are closed in sequence with priority. In extremely poor cases, small components that cannot meet any matching conditions are directly assigned to the defective product tray, realizing automatic sorting of good and defective products.

[0011] Automation and production line integration The component selection schemes generated by AI algorithms are transformed into control commands for automated equipment. Through automated production lines, the automatic selection, feeding, and assembly of components are realized, completing the entire process from AI selection to actual assembly and achieving deep integration of factory digital systems, AI algorithms, and automation systems.

[0012] Real-time dynamic optimization and adjustment The system collects real-time data on the distribution of products with various torque values ​​during the production process, dynamically assesses changes in data distribution, and promptly adjusts the placement of materials in the tray to complete full tray placement as quickly as possible. Based on the data from the previous batch of production, a preliminary torque value matching scheme is formulated. The AI ​​matching model and matching scheme are continuously optimized by combining the real-time production data of the day, with the set range as the precise target for tray placement, to achieve dynamic iterative optimization of torque matching.

[0013] The beneficial effects of this invention are: Significantly improve torque consistency: Through AI-powered precise selection, the torque tolerance of the entire shaft after selecting the hinge components is controlled within 20N, solving the problems of large torque fluctuations and wide torque range of the entire shaft in existing technologies, and significantly improving the feel consistency of the phone hinge.

[0014] Improve production yield: Achieve optimal matching of small components through AI matching rules, reduce defective products caused by component mismatch, and significantly improve overall production yield by combining with automatic sorting of defective products.

[0015] Achieve full-process intelligence and automation: Deeply integrate AI algorithms, MES digital systems and automated production lines to achieve full-process automation and intelligence in torque data acquisition, analysis, selection, tray placement and assembly, replacing traditional manual pairing and improving production efficiency.

[0016] Full-process data traceability: Through measures such as QR code laser engraving, 1:1 data binding, and foolproof trays, full-process traceability of torque data of small components is achieved, which facilitates quality control and problem investigation during the production process.

[0017] Dynamic optimization of matching scheme: Based on real-time production data, the data distribution is dynamically evaluated, and the AI ​​matching model and tray placement scheme are continuously optimized to adapt the matching scheme to the dynamic changes in the production process, ensuring matching accuracy and efficiency. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a mobile phone hinge torque matching method proposed in this invention; Figure 2 This is a flowchart illustrating the AI ​​pairing and automated combination process of a mobile phone hinge torque matching method proposed in this invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0020] refer to Figures 1-2 A method for matching the torque of a mobile phone hinge includes the following steps: Collect engineering design data and preprocess it to generate a structured engineering design data set; Perform entity recognition and relation extraction operations on the structured engineering design dataset to generate a set of engineering entity nodes and a set of relation edges, and construct an engineering design knowledge graph and entity relation graph; Based on the constraint rule information in the structured engineering design dataset, a set of constraint nodes and a set of constraint dependencies are constructed to generate a constraint dependency graph; Receive engineering design knowledge retrieval requests, perform semantic parsing operations on the engineering design knowledge retrieval requests, extract design object information, design target information and design constraint information, generate design object vectors, design target vectors and design constraint vectors, and map the design object vectors to the set of starting entity nodes in the entity relationship graph; The initial set of entity nodes, the design target vector, and the design constraint vector are input into the improved NBFNet model. Candidate domain expansion propagation calculation is performed in the entity relationship graph to generate an initial candidate node set. The initial candidate node set is mapped to the constraint dependency graph to perform constraint coupling propagation calculation to generate a constraint-satisfied node set. The constraint-satisfied node set is mapped back to the entity relationship graph to perform target convergence propagation calculation under the guidance of the design target vector to generate a target-matching candidate node set. During the propagation process, the relationship propagation state is updated through a relationship memory gating structure. Perform path consistency calculation on the inference path corresponding to the target matching candidate node set, and generate a path score set based on path structure coherence, design object matching degree, design target matching degree and design constraint satisfaction degree; Based on the path score set, the target matching candidate node set is sorted to generate an engineering design knowledge retrieval result set and output the corresponding reasoning path.

[0021] In this embodiment, the engineering design data includes engineering design document data, engineering specification text data, historical design scheme data, equipment parameter data, and component description data. The preprocessing steps include performing text parsing on the engineering design data to generate a text sequence set and a parameter field set; performing data cleaning on the text sequence set and parameter field set to delete duplicate records and unify field formats; performing terminology standardization on the text sequence set to map engineering terms to standard engineering terminology identifiers; and performing field structuring on the text sequence set and parameter field set to generate a structured engineering design data set.

[0022] In this embodiment, the steps for performing entity recognition and relation extraction operations include: Example 1:

[0023] In this embodiment, a total of 1886 pieces of torque data for the mobile phone hinge component were collected. After preprocessing, four invalid data points in the low range of 83-86N were removed, leaving 1882 pieces of valid data. The Anderson-Darling normality test showed that the mean of this batch of data was 126.22N, the median was 126.80N, and the standard deviation was 7.69N. The majority of the data were concentrated in the 110-139N range.

[0024] An AI torque matching model was constructed based on a neural network algorithm. After training, the median of the data distribution was calculated, and the target torque range of the shaft was set to 250±10N (240-260N). A multi-tray priority matching mechanism was adopted, with matching tray 1 as the first priority. 1882 pieces of valid data were paired in pairs. Through the pairing logic of "smaller number to larger number", 941 optimal pairs were achieved, with a yield rate of 99.78%. The torque values ​​of all paired shafts fell within the 240-260N range, and the torque tolerance was controlled within 20N.

[0025] The AI-based selection and matching scheme is transformed into control commands for automated equipment. Through automated production lines, small components are automatically selected and assembled. At the same time, production data is collected in real time, and the placement of materials on the trays is dynamically adjusted, improving the efficiency of full tray loading by more than 30%. Example 2:

[0026] In this embodiment, a total of 144 PCS (Group 2) of torque data for mobile phone hinge components produced on the same day were collected. The minimum value was 103.7N and the maximum value was 135.9N. Based on the trained AI matching model, the optimal matching range was set to 250±10N (240-260N), and the suboptimal ranges were 250±20N (230-270N) and 250±30N (220-280N).

[0027] The first step involved completing 32 optimal pairings within the 240-260N range, using 64 pieces of data, achieving a first-pass yield of 44.4%. The second step involved completing 25 suboptimal pairings within the 230-270N range using the remaining 80 pieces of data, using 50 pieces of data, achieving a first-pass yield of 34.72% and a cumulative yield of 79.2%. The third step involved completing 15 pairings within the 220-280N range using the remaining 30 pieces of data, using 30 pieces of data, achieving a first-pass yield of 20.83% and a cumulative yield of 100%.

[0028] This graded pairing method enables the effective use of all data, leaving no surplus good data. At the same time, the assembly is completed through an automated production line, and the consistency of the shaft torque feel is improved by more than 80% compared with traditional manual pairing.

[0029] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for matching the torque of a mobile phone hinge, characterized in that, Includes the following steps: Torsional mechanical data of the mobile phone hinge component were collected. After preprocessing the data, mechanical intervals were divided. The data volume of each interval was counted and normality test was performed to determine the valid and invalid data. An AI torque component selection model is built based on machine learning algorithms. The torque characteristics of small components are used as input and the predicted torque of the whole shaft is used as output. The model parameters are trained and optimized to improve the prediction accuracy and robustness of the model. The torque test results of the small components are uploaded to the factory's MES system, and the torque data of the small components is uniquely bound to the physical entity through a QR code. The trained AI model is used to calculate the median of the historical torque data distribution, and the target torque range of the whole axle is set. Based on this range, a pairwise pairing rule for small components is formulated, that is, the sum of the torque values ​​of two small components within the target torque range is the optimal pairing. AI-automated selection and tray placement is executed based on pairing rules. The multi-tray priority pairing mechanism completes the matching, pairing, and defective product sorting of small components. At the same time, it combines with automated equipment to realize the instruction conversion and automatic assembly of the selection scheme. Real-time acquisition of torque data during the production process, dynamic evaluation of data distribution, and real-time optimization of AI pairing model and tray arrangement scheme to achieve dynamic adjustment of torque matching.

2. The method for matching the torque of a mobile phone hinge according to claim 1, characterized in that, The data preprocessing includes removing invalid data. Invalid data refers to low-range data where the sum of any two pairs of torque values ​​does not fall within the set target torque range for the complete shaft. Valid data is the remaining data after deducting invalid data from the total data volume.

3. The method for matching the torque of a mobile phone hinge according to claim 1, characterized in that, The machine learning algorithm includes at least one of neural networks and support vector machines. Interference factors in production are incorporated into the model optimization process to enhance the algorithm's adaptability to actual production scenarios.

4. The method for matching the torque of a mobile phone hinge according to claim 1, characterized in that, The multi-tray priority pairing mechanism is as follows: with the pairing tray 1 as the first priority, small components that meet the optimal pairing conditions are first filled into tray 1. After the waiting queue of tray 1 is full, small components that do not meet the pairing conditions are sequentially transferred to the waiting queues of tray 2 and tray 3. After tray 1 is filled and then unloaded, trays 2 and 3 are unloaded in sequence with priority. In extremely poor cases, small components that cannot meet the matching conditions are directly assigned to the defective tray.

5. The method for matching the torque of a mobile phone hinge according to claim 1, characterized in that, The material trays of the optional display tray are set to an even-numbered column structure. The queue to be matched and the matching column are paired with each other from left to right using AI. The matching is completed when the sum of the torque values ​​of the small components in the two adjacent columns meets the mechanical specifications set by AI.

6. The method for matching the torque of a mobile phone hinge according to claim 1, characterized in that, The dynamic optimization includes formulating a preliminary pairing plan based on the previous batch of production data, adjusting the placement of materials in the pallet in combination with the real-time production data of the day, and using the target torque of the whole shaft as the precise target for pallet placement to complete the full pallet placement as quickly as possible.

7. A method for matching the torque of a mobile phone hinge according to claim 6, characterized in that, The method also includes process optimization of the automated assembly line for mobile phone hinge components. In processes such as component testing and loading, break-in, torque and angle testing, and sorting and tray placement, laser-engraved QR codes, tray error prevention, and 1:1 data binding measures are added to achieve full-process traceability of component data.

Images