Fiber-optic gyroscope assembly efficiency evaluation method based on digital twinning

Abstract

The invention relates to a fiber-optic gyroscope assembly efficiency evaluation method based on digital twinning, which aims at the working conditions of assembly workers and assembly equipment in the assembly process of a fiber-optic gyroscope, explores the efficiency evaluation method of a digital twinning technology in the field of man-machine cooperative assembly, and provides a method for evaluating the assembly efficiency of a fiber-optic gyroscope based on real-time updated man-machine interaction data in a digital twinning body. The method comprises the following steps: establishing a digital twinborn model for fiber-optic gyroscope assembly to perform basic operator assembly data acquisition, establishing a fiber-optic gyroscope assembly efficiency evaluation system from four aspects of personnel load, equipment level, man-machine interaction and assembly efficiency, and determining a calculation method of each index by calculating quantitative indexes and adopting technologies and theories such as motion capture and information entropy. Index efficiency data calculation based on digital twinning is realized, and a BP neural network model is introduced to carry out comprehensive evaluation on the assembly efficiency of the fiber-optic gyroscope.

Description

Technical Field

[0001] This invention relates to a method for evaluating the assembly performance of fiber optic gyroscopes based on digital twins, belonging to the field of advanced manufacturing technology. Background Technology

[0002] Fiber optic gyroscopes, as a core component of inertial navigation instruments, have wide applications in aerospace, shipbuilding, and military fields. The assembly of fiber optic gyroscopes significantly impacts their performance. Currently, assembly relies primarily on manual labor, making it difficult to guarantee a high pass rate. Introducing robots to assist workers in assembly is a crucial means to achieve intelligent manufacturing transformation of fiber optic gyroscopes. Digital twin technology, based on data acquisition and motion recognition, achieves "virtual-real fusion, using the virtual to control the real" between the digital twin and the physical entity. Digital twin technology can efficiently monitor the status of assembly scenarios in manufacturing and facilitate human-machine interaction research. Evaluating the efficiency of the fiber optic gyroscope assembly process based on digital twin technology is of great significance for accelerating the intelligent transformation of fiber optic gyroscopes and improving assembly quality.

[0003] Current research on the assembly process of fiber optic gyroscopes has yielded significant results. Sun Yuan et al. proposed a fiber dynamics modeling and simulation method based on discrete differential geometry theory for fiber optic pigtail assembly of fiber optic gyroscopes, providing theoretical support for low-stress assembly and path planning of the gyroscope optical path. Liu Haoting et al. proposed an endoscope system design and related image analysis method for quality control of interferometric fiber optic gyroscope (IFOG) assembly. They designed an image enhancement method based on the db4 wavelet basis, an image texture feature estimation method based on the gray-level co-occurrence matrix, and a method for calculating image geometric features, for analysis of endoscopic images. Liu Yuanyuan et al. considered the errors generated during the assembly of a three-axis integrated fiber optic gyroscope assembly with a shared light source. Based on the relationship between the light source power and the injection current, and the technical level of existing optical devices, they established optical path loss inequality constraints, defining the areas where product debugging was successful or not, and identifying and resolving components with abnormal optical path loss early. Fan Yunqiang et al. addressed the Shupe error, a major technical bottleneck in the engineering application of high-precision fiber optic gyroscopes, and experimentally revealed the influence mechanism of interferometric optical path assembly stress on the Shupe error. Zhang Bo et al., addressing the lightweight requirements of a certain type of fiber optic gyroscope, based on a frame-type body structure, adopted a three-axis integrated and optical path multiplexing scheme, achieving the design of a miniaturized three-axis integrated gyroscope through device miniaturization and multi-channel integration of gyroscope circuits. With in-depth research, digital twin technology has also played an important role in assembly process monitoring, human-computer interaction, and other fields. Zhu et al. studied a dynamic reconfiguration optimization method for intelligent manufacturing systems considering human-robot collaborative operations in a digital twin environment. Sun Huibin et al. studied digital twin-driven aero-engine assembly technology, discussing its composition, function, and process, and providing detailed analysis of key technology examples verifying assembly process control, parts selection, assembly operation guidance, assembly gap control, assembly technology status control, and assembly test data correlation analysis.

[0004] Based on the characteristics of existing fiber optic gyroscope assembly research, the shortcomings of existing technologies are summarized as follows:

[0005] (1) Currently, research on fiber optic gyroscope assembly often focuses on the technology and performance of the gyroscope itself, while the impact of assembly workers and equipment used in the assembly process is relatively limited. Existing research mainly explores the structural optimization, accuracy improvement, and optical, mechanical, and electronic systems involved in the assembly process of fiber optic gyroscopes, but there is little research on the role, operation behavior, and equipment used by assembly workers in actual assembly.

[0006] (2) Currently, the application of digital twin technology in the assembly process of fiber optic gyroscopes is relatively limited. This makes it difficult to achieve real-time status monitoring, accurate data acquisition, and dynamic analysis of the assembly process, thus affecting assembly efficiency and quality assurance. Summary of the Invention

[0007] The technical problem solved by this invention is to address the technical deficiencies in the assembly of fiber optic gyroscopes in the existing technology by proposing a method for evaluating the assembly performance of fiber optic gyroscopes based on digital twins.

[0008] The present invention solves the above-mentioned technical problem through the following technical solution:

[0009] A method for evaluating the assembly performance of fiber optic gyroscopes based on digital twins, comprising:

[0010] Establish evaluation criteria for the assembly performance of fiber optic gyroscopes and determine the corresponding quantitative indicators for each evaluation criterion.

[0011] Establish a digital twin model of the fiber optic gyroscope assembly process;

[0012] The assembly data of basic operators for various fiber optic gyroscope assembly performance evaluation standards were collected using digital twin models.

[0013] Quantitative indicators are calculated based on the collected data to obtain the performance values ​​of each quantitative indicator.

[0014] A neural network model was established and trained to obtain an evaluation model for the assembly performance of fiber optic gyroscopes. The performance values ​​of various quantitative indicators were used as input data to obtain the evaluation results of the assembly performance of fiber optic gyroscopes.

[0015] The evaluation criteria for the assembly efficiency of the fiber optic gyroscope include personnel workload, equipment level, human-computer interaction, and assembly efficiency, among which:

[0016] The quantitative indicators of personnel workload are: upper limb task load, mobility task load, operational error rate, and reaction speed;

[0017] The quantitative indicators of equipment level are: degree of automation, degree of digitalization, and equipment failure rate;

[0018] The quantitative indicators of human-computer interaction are: logical complexity, operational complexity, and degree of logistics automation.

[0019] The quantitative indicator for assembly efficiency is: production efficiency.

[0020] The digital twin model is established based on the fiber optic gyroscope assembly scenario, operator, and equipment information. It is used to collect basic operator assembly data corresponding to each quantitative indicator for quantitative indicator calculation.

[0021] The method for calculating the quantitative indicator effectiveness value of the personnel workload is as follows:

[0022] Upper limb workload u 11 :

[0023]

[0024] In the formula, This represents the variance of the force at the elbow joint during upper limb movement. This represents the variance of the torque at the elbow joint during upper limb movement.

[0025] Mobile task load u 12 :

[0026]

[0027] In the formula, M is the load of the assembly worker, g is the acceleration per kilometer, and |h t | represents the absolute value of the vertical movement of the assembly worker's center of gravity;

[0028] Operational error rate u 13 :

[0029]

[0030] In the formula, n + This represents the number of instructions correctly judged by the assembly equipment within a certain period of time, and N represents the total number of instructions issued by people within that period of time.

[0031] Response sensitivity u 14 :

[0032]

[0033] In the formula, t1, t2, ..., t n The delay data for each event is represented by the average value, which characterizes the sensitivity of the system.

[0034] The variance of the force at the elbow joint during upper limb movement Variance of torque at the elbow joint during upper limb movement The calculation method is as follows:

[0035]

[0036] In the formula, F1 is the joint force on the first joint, T1 is the joint torque on the first joint, and F i T represents the joint force acting on the i-th joint. i Let F be the joint torque acting on the i-th joint, G1 and G2 be the external load weight and the weight of the forearm, respectively. g F m T g T m F and T represent the active force, passive force, active torque, and passive torque borne by the current joint drive segment, respectively. F and T are the total joint torques.

[0037] The method for calculating the effectiveness value, a quantitative indicator of the equipment level, is as follows:

[0038] Automation level u 21 :

[0039]

[0040] In the formula, n A1 n represents the number of tasks completed automatically. A Indicates the total number of tasks;

[0041]

[0042] In the formula, n A2 n represents the number of processes requiring human intervention. T This indicates the total number of processes involved in the production of a single product.

[0043] u 21 =0.6A1 + 0.4A2

[0044] Level of digitalization:

[0045]

[0046] In the formula, n D N represents the number of data points collected. D This indicates the theoretically collectable number of data points;

[0047] Equipment failure rate u 23 :

[0048]

[0049] In the formula, n b N represents the number of time periods during which a failure occurred. O This indicates the total number of time periods during which the equipment operates.

[0050] The calculation method for the quantitative performance indicators of the human-computer interaction is as follows:

[0051] Logical complexity u 31 :

[0052] u 31 =H L

[0053] In the formula, H L The entropy value of the human-computer interaction process;

[0054] Operational complexity u 32 :

[0055] u 32 =H O

[0056] In the formula, H OThe second-order entropy value in the human-computer interaction process;

[0057] Logistics automation level u 33 :

[0058]

[0059] In the formula, n T N represents the number of transportation tasks completed automatically. T This indicates the total number of transportation tasks.

[0060] The first-order entropy value H of the human-computer interaction process L The calculation method is as follows:

[0061]

[0062] In the formula, n is the number of node types in PCG, p(A i ) represents the node i belonging to the Ath node. i The probability of a node of a certain type;

[0063] The second entropy value H of the human-computer interaction process O The calculation method is as follows:

[0064]

[0065] In the process of calculating the first-order entropy value, the classification of nodes is based on the human-computer interaction process in the digital twin model. In the program control flow (PCG), the actions of humans or machines are used as nodes, and the information interaction between different actions is used as the connection to achieve node selection.

[0066] During the second-order entropy calculation, the nodes are grouped according to the names of different steps in the program control process (CLP) of the digital twin model. When the names of adjacent nodes are different, the number of groups is different and the second-order entropy calculation results are different.

[0067] The calculation method for the quantitative indicator of assembly efficiency is as follows:

[0068] Production efficiency u 41 :

[0069]

[0070] In the formula, t is the total time to complete all tasks, and n is the number of tasks.

[0071] The neural network model is configured with 11 neurons in the input layer, 1 neuron in the output layer, and 64 neurons in the hidden layer. Historical usage data on the evaluation of the fiber optic gyroscope assembly process is used as the training set for the neural network model.

[0072] Twenty models were selected for training. 50% of the evaluation data was used as the training set, and the remaining 50% as the test set. All data was normalized to obtain the trained fiber optic gyroscope assembly performance evaluation model. The normalization method was as follows:

[0073]

[0074] In the formula, y i x is the normalized index value. i x represents the original index value. min and x max These represent the minimum and maximum values ​​of the quantitative indicator in the historical dataset, respectively.

[0075] The advantages of this invention compared to the prior art are:

[0076] (1) This invention provides a method for evaluating the assembly efficiency of fiber optic gyroscopes based on digital twins. A digital twin model is established for the assembly process, and through deep integration of virtual and real data, accurate data collection of the assembly process is achieved. This ensures that various data from the assembly process can be acquired in a timely, efficient, and accurate manner, providing a reliable data foundation for the calculation of multi-dimensional evaluation indicators. This, in turn, guarantees the accuracy of the evaluation results and the scientific nature of the decision-making. Simultaneously, the assembly efficiency is evaluated from four aspects: personnel workload, equipment level, human-computer interaction, and assembly efficiency. This achieves multi-level and multi-angle quantitative measurement of the assembly process, taking into account safety, work quality, production efficiency, and labor intensity. This is beneficial for enterprises to achieve refined management of the assembly process.

[0077] (2) This invention employs a neural network model to achieve intelligent, data-driven comprehensive evaluation of assembly efficiency based on different evaluation indicators. It effectively integrates information from various evaluation indicators such as load, cognition, and efficiency, providing a precise and comprehensive evaluation of assembly efficiency. The specific implementation of the method is demonstrated by showcasing an efficiency assessment of a fiber optic gyroscope assembly workshop in a factory. Simultaneously, a quantitative indicator model is used to calculate relevant indicators for various aspects of past assembly processes in the workshop. Furthermore, multiple experts score the assembly process, resulting in an objective evaluation of the assembly process efficiency. Attached Figure Description

[0078] Figure 1 A schematic diagram of the fiber optic gyroscope assembly performance evaluation system based on digital twin provided by the present invention;

[0079] Figure 2 A schematic diagram of the assembly worker's digital twin provided by the present invention;

[0080] Figure 3 A schematic diagram of a ball-and-stick model in a human digital twin provided by the present invention;

[0081] Figure 4 A schematic diagram of the assembly process PCG provided for this invention;

[0082] Figure 5 The BP neural network training flowchart provided by this invention;

[0083] Figure 6 This is a schematic diagram of the neural network training results provided by the present invention;

[0084] Figure 7 This is a schematic diagram illustrating the convergence of neural network training provided by the present invention. Detailed Implementation

[0085] This paper presents a digital twin-based method for evaluating the assembly efficiency of fiber optic gyroscopes. It explores the application of digital twin technology in human-machine collaborative assembly, focusing on the work of assembly workers and equipment during the assembly process. Based on real-time updated human-machine interaction data within the digital twin, a digital twin model of the fiber optic gyroscope assembly is established for basic operator assembly data collection. An evaluation system for fiber optic gyroscope assembly efficiency is established from four aspects: personnel workload, equipment level, human-machine interaction, and assembly efficiency. By calculating and quantifying indicators, the calculation methods for each indicator are determined using motion capture, information entropy, and other technologies and theories. This enables the calculation of indicator efficiency data based on digital twins, and a BP neural network model is introduced for comprehensive evaluation of fiber optic gyroscope assembly efficiency.

[0086] A method for evaluating the assembly performance of fiber optic gyroscopes based on digital twins, such as Figure 5 As shown, the steps are as follows:

[0087] Establish evaluation criteria for the assembly performance of fiber optic gyroscopes and determine the corresponding quantitative indicators for each evaluation criterion.

[0088] Establish a digital twin model of the fiber optic gyroscope assembly process;

[0089] The assembly data of basic operators for various fiber optic gyroscope assembly performance evaluation standards were collected using digital twin models.

[0090] Quantitative indicators are calculated based on the collected data to obtain the performance values ​​of each quantitative indicator.

[0091] A neural network model was established and trained to obtain an evaluation model for the assembly performance of fiber optic gyroscopes. The performance values ​​of various quantitative indicators were used as input data to obtain the evaluation results of the assembly performance of fiber optic gyroscopes.

[0092] The evaluation criteria for fiber optic gyroscope assembly efficiency include personnel workload, equipment level, human-machine interaction, and assembly efficiency, among which:

[0093] The quantitative indicators of personnel workload are: upper limb task load, mobility task load, operational error rate, and reaction speed;

[0094] The quantitative indicators of equipment level are: degree of automation, degree of digitalization, and equipment failure rate;

[0095] The quantitative indicators of human-computer interaction are: logical complexity, operational complexity, and degree of logistics automation.

[0096] The quantitative indicator for assembly efficiency is: production efficiency.

[0097] The digital twin model is established based on the fiber optic gyroscope assembly scenario, operator, and equipment information. It is used to collect basic operator assembly data corresponding to each quantitative indicator for quantitative indicator calculation.

[0098] The neural network model is set with 11 neurons in the input layer, 1 neuron in the output layer, and 64 neurons in the hidden layer. Historical usage data on the evaluation of the fiber optic gyroscope assembly process is used as the training set for the neural network model.

[0099] Twenty model training sets were selected, with 50% of the evaluation data used as the training set and 50% as the test set. After normalizing all the data, the trained fiber optic gyroscope assembly performance evaluation model was obtained.

[0100] The following description, in conjunction with the accompanying drawings and preferred embodiments, provides further details:

[0101] In the current embodiment, while promoting the digitalization and intelligentization of the fiber optic gyroscope assembly process and employing technologies such as digital twins and robotics, a highly efficient and accurate performance evaluation system suitable for new assembly scenarios is lacking. Addressing the issue of constructing a performance evaluation system for fiber optic gyroscope assembly scenarios integrating new technologies, this paper, based on the widely used digital twin model in the current manufacturing informatization process, and after a comprehensive and in-depth analysis of the pain points in performance evaluation of human-machine collaborative assembly processes and the characteristics of digital twin technology, establishes a set of fiber optic gyroscope assembly performance evaluation index systems based on digital twins from four aspects: personnel workload, equipment level, human-machine interaction, and assembly efficiency. This system guides the optimization of the fiber optic gyroscope assembly process and the improvement of key aspects, such as... Figure 1 As shown.

[0102] The evaluation criteria for fiber optic gyroscope assembly efficiency include personnel workload, equipment level, human-machine interaction, and assembly efficiency, specifically:

[0103] I. Establishment of Evaluation Criteria for Fiber Optic Gyroscope Assembly Performance

[0104] (1) Staff load

[0105] The evaluation of fiber optic gyroscope assembly performance mainly studies the human operation and interaction methods with the machine during the assembly process. The workload generated by the assembly personnel's interaction with the objects (fiber optic gyroscope components and assembly equipment) is a key factor affecting work quality; therefore, evaluating fiber optic gyroscope assembly performance from the perspective of personnel workload is necessary. By conducting motion capture and biomechanical analysis on the assembly personnel, and using digital twin technology to create digital twins of the assembly personnel, such as... Figure 2 As shown, the evaluation will be conducted from the following two aspects:

[0106] Traditional task load assessment methods often rely on extensive experiments and surveys to determine the workload of a task. This process involves numerous factors and the results are somewhat subjective, making it difficult to achieve a qualitative and quantitative assessment of task load. Digital twins of assembly personnel, however, can model, capture, and synchronize the personnel's bodies. Further computer-aided analysis can then provide a qualitative and quantitative assessment of the workload of assembly personnel during the fiber optic gyroscope assembly process. Subsequent steps can be performed after the virtual-real mapping between the personnel and the digital twin is established.

[0107] The cognitive load of assemblers during fiber optic gyroscope assembly stems from their acquisition, understanding, processing, and reaction to information. This requires the integrated use of their senses, memory, and thinking abilities, resulting in a certain level of cognitive load. Maintaining the cognitive load of assemblers within a scientifically reasonable range ensures the safety and efficiency of the operation. Conversely, a cognitive load exceeding the normal range consumes more physical and mental energy, negatively impacting the fiber optic gyroscope assembly process. A digital twin model of the fiber optic gyroscope assembly process includes performance indicators for many interactive processes. By comprehensively analyzing relevant parameters in the twin data, the intensity of human cognitive load during fiber optic gyroscope assembly can be reflected.

[0108] (2) Equipment level

[0109] In a digital twin-based assembly efficiency evaluation system, equipment level is a key dimension for measuring the performance and technical capabilities of equipment in the assembly process. It comprehensively reflects the overall performance of assembly equipment in terms of technical capabilities, operational status, and information interaction. The level of equipment has a significant impact on the assembly process, resource utilization, and system reliability. Equipment level can be further subdivided into two aspects: equipment intelligence level and equipment quality. Simultaneously, equipment level also has a significant impact on the efficiency and accuracy of the application of digital twin technology in the assembly process and data acquisition.

[0110] The level of equipment intelligence refers to the ability of equipment to utilize automation and digitalization technologies during the assembly process, and is an important component describing the level of equipment in new assembly scenarios. Improving the level of intelligence can effectively increase assembly efficiency, reduce manual intervention, and enhance the flexibility and adaptability of the system, thereby meeting complex and personalized assembly needs. The degree of automation reflects the equipment's ability to autonomously perform tasks during the assembly process. A high degree of automation can significantly improve assembly efficiency, reduce the labor intensity of assembly workers, and make the assembly process more controllable. The equipment's ability to acquire, process, and interact with information is an important indicator for evaluating the deep integration of equipment with digital twin systems. Highly digitized equipment can collect, transmit, and analyze key data during the assembly process in real time, providing accurate and dynamic input for the digital twin model, supporting simulation and optimization. Simultaneously, it can enhance the traceability and flexible manufacturing capabilities of the assembly process, meeting personalized and complex assembly needs.

[0111] Equipment quality refers to the comprehensive characteristics of equipment during operation, including stability, reliability, and service life. Equipment is the foundation for the stable operation of an assembly system. Reliable equipment quality reduces unplanned downtime, ensures efficient assembly processes and stable execution of production plans, and guarantees the efficient production of high-quality products, directly impacting the continuity of assembly tasks and the achievement of production targets. Furthermore, improved equipment quality reduces maintenance rates and per-maintenance costs, thereby enhancing the overall economic efficiency of the assembly system.

[0112] (3) Human-computer interaction

[0113] In the assembly process of fiber optic gyroscopes, the interaction between assembly workers, assembly equipment, and the assembled components directly affects the degree of task completion. Human-machine interaction needs to consider the complexity of the assembly workers' operation of the equipment during assembly, as well as the logical complexity of the interaction process. Meanwhile, the degree of automation in logistics is also an important indicator for measuring the human-machine interaction performance of the assembly system.

[0114] In traditional manufacturing processes, assembly personnel primarily perform assembly operations using simple tools. With the introduction of technologies such as intelligent robots, their roles have gradually shifted to control, supervision, and collaboration. Humans expend less physical energy during assembly, allowing them to better understand the interaction logic. Therefore, evaluating the logical complexity of the fiber optic gyroscope assembly process can significantly aid in designing a rational assembly workflow and enhancing auxiliary capabilities. The complexity evaluation of the fiber optic gyroscope assembly process based on digital twins includes assessing the complexity of the interaction's inherent logical structure, the complexity of the actions required to complete the assembly process, and the level of knowledge and cognitive capacity required for fiber optic gyroscope assembly.

[0115] In traditional manufacturing processes, operators primarily operate machine tools and other equipment. With the introduction of human-computer interaction (HCI), their roles have gradually shifted to control, supervision, and collaboration. Humans expend less physical energy during operation, leading to greater understanding of the interaction logic. Therefore, evaluating the logical complexity of HCI processes can significantly aid in designing reasonable interaction flows and enhancing assisted learning. The complexity assessment of HCI processes based on digital twins includes evaluating the complexity of the interaction's inherent logical structure, the complexity of the actions required to complete the HCI process, and the level of knowledge and cognitive capacity required for HCI interaction.

[0116] The level of automation in logistics directly impacts the interaction between assembly workers and logistics equipment. Logistics equipment with low levels of automation often requires workers to perform more complex operations during loading and unloading. The way assembly workers load and unload materials and move logistics equipment also affects the physical workload and difficulty of their work. As logistics automation increases, the interaction between assembly workers and logistics equipment shifts from execution to monitoring and intelligence, significantly improving the depth and efficiency of the interaction.

[0117] (4) Assembly efficiency

[0118] The goal of fiber optic gyroscope assembly is to help assemblers complete tasks more safely, reliably, and efficiently through information exchange between humans and equipment. Therefore, the efficiency of fiber optic gyroscope assembly should be evaluated from an ergonomic perspective.

[0119] The fiber optic gyroscope assembly process aims to improve existing assembly procedures, reduce the workload of assembly personnel, and increase production efficiency. Evaluating production efficiency is the most direct reflection of the fiber optic gyroscope assembly effect. Twin data contains a series of recorded data from the production process, facilitating the calculation of production efficiency.

[0120] (1-1) The quantitative indicators of personnel workload are: upper limb task load, mobile task load, operational error rate, and reaction sensitivity;

[0121] During the assembly process, assembly workers experience upper limb workload from performing assembly tasks using their upper limbs, as well as movement workload from moving between various workbenches and storage areas. The workload encompasses both of these aspects.

[0122] Considering the characteristics of the fiber optic gyroscope assembly process and feedback from frontline assembly workers, excessive physical exertion can lead to fatigue and even threaten work safety. Investigations revealed that injuries caused by excessive physical exertion are primarily concentrated in the arms and lumbar spine. Therefore, this paper uses the digital twin of the assembly worker as a representative example to evaluate the upper limb physical exertion intensity during the task.

[0123] First, based on data transmission technology, a digital twin of the assembly worker is constructed. This digital twin is treated as a multi-rigid-body model, and a ball-and-stick model is further used to represent its arm portion, such as... Figure 3 As shown.

[0124] Furthermore, according to d'Alembert's principle, the forces and moments that a joint's reaction force and reaction torque need to resist include the active forces and torques caused by gravity and load, and the passive forces and passive torques generated by the movement, that is:

[0125]

[0126] Where F represents the joint force borne by a certain joint, and T represents the joint torque borne by a certain joint. g F m T g T m The active force, passive force, active torque, and passive torque borne by the drive segment of this joint.

[0127] In the ball-and-stick model, relevant parameters of the upper limbs are defined, such as... Figure 3 As shown, A, B, and C correspond to the center points of the human shoulder joint, elbow joint, and hand in physical space, respectively, and their spatial coordinates can be obtained in real time in the human digital twin. G1 and G2 represent the external load weight and the weight of the human forearm, respectively.

[0128] The vertical angle θ at the elbow joint is easily obtained as follows:

[0129]

[0130] The angular acceleration α is:

[0131]

[0132] Simplifying the upper limb movement to a motion where the upper arm remains stationary while the forearm rotates around the elbow joint as a fulcrum, we can obtain the joint force and torque at the elbow joint as follows:

[0133]

[0134] Variance can characterize the stability of a series of values ​​and reflect the degree of fluctuation. Considering that the stability of the elbow joint decreases when assembly workers are under heavy workloads, leading to fluctuations in joint force and torque at the elbow joint, it is advisable to select an appropriate time window, calculate the variance, and then perform a comprehensive weighted average to characterize the upper limb workload of assembly workers.

[0135]

[0136] At the assembly station, assembly workers need to move to specific positions for assembly. A digital twin model of the assembly worker's center of gravity movement trajectory is established. The load on the assembly worker is denoted as M, including their own weight and the weight of the object being moved. To simplify the model, the assembly worker and their load are abstracted as a center of gravity point, and the work done during the movement of the center of gravity is recorded. Since gravity is always vertically downward, no work is done in the horizontal direction; therefore, the change in the vertical coordinate h during the movement of the center of gravity is recorded. The work done by the assembly worker during the movement characterizes the workload generated by the movement task.

[0137] ΔW=Mg|Δh| (8)

[0138]

[0139] u 12 =W (10)

[0140] Cognitive load is a multidimensional concept that is related to the work requirements, time pressure, operator ability and effort of the assemblers during the fiber optic gyroscope assembly process. Therefore, exploring the cognitive load of the assemblers during the fiber optic gyroscope assembly process can reflect the degree to which a person's cyber-physical system is used.

[0141] Based on the characteristics of the digital twin model of fiber optic gyroscope assembly, we consider using behavioral performance measurement methods to measure and evaluate the cognitive load of assembly personnel, mainly including two performance factors: operational error rate and reaction sensitivity.

[0142] The actions of the assembly personnel are captured, recognized, and converted, ultimately obtaining feedback from the machine to form a complete fiber optic gyroscope assembly process. This entire process can be recorded in digital twin data and displayed through a synchronized digital twin model. The probability of errors occurring during the fiber optic gyroscope assembly process reflects the accuracy of the assembly equipment's response to human actions and commands.

[0143]

[0144] Where, n + This represents the number of instructions correctly determined by the assembly equipment within a certain period, and N represents the total number of instructions issued by people within that period.

[0145] The interactive sensitivity of the fiber optic gyroscope assembly process refers to the responsiveness of a human to the actions, directions, and other information of the assembly equipment. This is achieved by monitoring the human-machine interaction event logs during the fiber optic gyroscope assembly process using a digital twin, which includes event trigger latency data. The sensitivity description of the human-machine interaction process will reflect the system's processing latency for each event.

[0146]

[0147] Where t1, t2, ..., t n The delay data for each event is represented by the average value, which characterizes the sensitivity of the system.

[0148] (2-1) The degree of automation of an assembly station is determined by both the automation rate and the frequency of human intervention. If the automation rate is high and the frequency of human intervention is low, it indicates that the automation level of the station is relatively high. If the automation rate is high but the frequency of human intervention is also high, it may be due to the complexity of the equipment leading to frequent debugging, or the environmental conditions not yet mature. If the automation rate is low and the frequency of human intervention is low, it indicates that the station is mainly operated manually and there is little application of automation.

[0149] The expression for the automation rate A1 is:

[0150]

[0151] Where, n A1 n represents the number of tasks completed automatically. A This indicates the total number of tasks.

[0152] The frequency of human intervention expresses the number of times human intervention is required in the assembly process of a product. The expression for the frequency of human intervention A2 is:

[0153]

[0154] Where n A2 n represents the number of processes requiring human intervention. T This indicates the total number of processes involved in the production of a single product.

[0155] Finally, these two factors are combined using a weighted method, as shown in the following expression:

[0156] u 21 =0.6A1+0.4A2 (15)

[0157] The degree of digitalization of assembly stations is determined by the data acquisition coverage rate. 23 The decision represents the percentage of data points that can be collected at an assembly station relative to the theoretical maximum number of data points that can be collected. Its expression is:

[0158]

[0159] Where, n D N represents the number of data points collected. D This indicates the theoretical number of data points that can be collected.

[0160] Equipment failure rate u 23 The expression describing the frequency of malfunctions or errors occurring during equipment operation is as follows:

[0161]

[0162] Where n b N represents the number of time periods during which a failure occurred. O This indicates the total number of time periods during which the equipment operates.

[0163] (3-1) The logical complexity of the assembly process of a fiber optic gyroscope based on digital twins mainly evaluates the complexity of the logical structure of the interaction between humans and the assembly equipment. The method of calculating the logical structure of the interaction itself using first-order entropy is based on a program control graph (PCG). First, the complete human-machine interaction process is represented using a PCG, where nodes represent human or machine actions in the twin data, and lines represent the information flow between different actions recorded in the twin data. The first-order entropy calculation method is as follows:

[0164]

[0165] Where n is the number of node types in PCG, p(A i ) represents a node belonging to the Ath node. i The probability of a class node.

[0166] An example of the PCG assembly process for fiber optic gyroscopes built using twin data is shown below. Figure 4 As shown.

[0167] Further considering the inputs and outputs of each node in the PCG, if some nodes have the same number of inputs and outputs, then these nodes are considered to have equal levels, that is, the same entropy value. For Figure 4 The node classifications in the above cases are shown in Table 1.

[0168] Table 1. Node Classification Table

[0169]

[0170] According to the formula, Figure 4 The entropy value of the human-computer interaction process is calculated as follows:

[0171]

[0172] After calculating the entropy value, a corresponding evaluation index of logical complexity is obtained based on the entropy value.

[0173] u 31 =H L

[0174] The operational complexity of the fiber optic gyroscope assembly process based on digital twins mainly evaluates the complexity of the assembly operation steps themselves. The method for calculating the assembly operation complexity using second-order entropy is still based on the Control Graph (CLP), and the process of constructing the CLP from the twin data is the same as above. The formula for calculating the second-order entropy is the same as equation (13), the difference being the classification method of the nodes: the grouping for calculating the second-order entropy is based on the different names of the adjacent nodes of each node in the CLP. The more similar nodes there are in the graph, the fewer the groups, the better the symmetry of the graph, and the smaller the second-order entropy. Figure 4 The second-order entropy groupings are shown in Table 2.

[0175] Table 2. Figure 4 Corresponding second-order entropy grouping

[0176]

[0177] According to formula (13), Figure 4 The second-order entropy value of the human-computer interaction process is calculated as follows:

[0178]

[0179] After calculating the entropy value, a corresponding evaluation index for operational complexity is obtained based on the entropy value.

[0180] u 32 =H O

[0181] The degree of logistics automation is mainly measured by the proportion of logistics transportation tasks in the assembly workshop that are completed by automated equipment (such as AGVs, AMRs, conveyor lines, etc.), i.e., the automated transportation coverage rate u. 33 The decision, expressed as follows:

[0182]

[0183] Where, n T N represents the number of transportation tasks completed automatically. T This indicates the total number of transportation tasks.

[0184] (4-1) Twin data includes parameters such as the working time and output of assembly workers. The production efficiency of assembly workers in the assembly process refers to the average time required to complete a task.

[0185]

[0186] u 41 The value directly reflects the improvement effect that fiber optic gyroscope assembly technology brings to the operator's production.

[0187] II. Comprehensive Evaluation of BP-Based Fiber Optic Gyroscope Assembly Performance

[0188] Backpropagation (BP) neural networks are a widely used artificial neural network model, mainly composed of an input layer, hidden layers, and an output layer. Each layer contains multiple neurons connected by weights. BP neural networks learn the mapping relationship between input and output data, continuously adjusting the connection weights to process the input data and generate the corresponding output. The core idea of ​​BP neural networks is to use the steepest descent learning rule to continuously adjust the network's weights and thresholds, thereby obtaining a network with the minimum sum of squared errors. This paper uses a typical 3-layer BP neural network to construct an evaluation model.

[0189] Based on the evaluation index system constructed above, the number of neurons in the input layer of the neural network is set to 11, the number of neurons in the output layer to 1, and the number of neurons in the hidden layer to 64. The evaluation of the fiber optic gyroscope assembly process based on historical usage data constitutes the model training set. The process of learning, training, and evaluating the BP neural network is as follows: Figure 6 As shown.

[0190] Example:

[0191] This paper demonstrates the specific implementation of the method by showcasing an efficiency evaluation of a fiber optic gyroscope assembly workshop in a factory. A quantitative indicator model is used to calculate various indicators related to the workshop's past assembly processes. Multiple experts then score the assembly process, resulting in an objective evaluation of its efficiency.

[0192] (1) Historical data on fiber optic gyroscope assembly is shown in Table 2. Each evaluation indicator is distinguished by a hierarchical numerical code. The expert score is the average of the scores obtained by multiple experts on a 100-point scale.

[0193] Table 2 Historical Data of Fiber Optic Gyroscope Assembly

[0194]

[0195]

[0196] The algorithm is implemented as follows:

[0197] The neural network has 11 neurons in the input layer, 1 neuron in the output layer, and 64 neurons in the hidden layers. The model training set is constructed based on evaluations of the fiber optic gyroscope assembly process using historical usage data.

[0198] This paper performs neural network computation on the given dataset of 20 fiber optic gyroscope assembly data, setting 50% of the evaluation data as the training set and 50% as the test set. To avoid the impact of differences in unit values ​​between different evaluation metrics on training accuracy, the historical workshop data is first normalized before training, as shown in the following formula:

[0199]

[0200] Among them, y i x is the normalized index value. i x represents the original index value. min and x max These are the minimum and maximum values ​​of the corresponding indicators in the historical dataset, respectively.

[0201] The final training results are as follows Figure 6 As shown, the neural network training convergence process is as follows: Figure 7 As shown, Figure 6 Each point in the graph corresponds one-to-one with the data in the test set. The horizontal axis represents the expert evaluation value, and the vertical axis represents the predicted value of the neural network model. The straight line in the graph represents the data where prediction = True Value, that is, the line formed when the predicted value obtained by the neural network is the same as the actual expert evaluation value. The goodness-of-fit R-squared is obtained by performing a goodness-of-fit analysis between the test set data points and the straight line. 2 =0.9955, which is close to 1, indicating that the fiber optic gyroscope assembly performance evaluation system designed in this paper has a good fit and can accurately reflect the matching relationship of various performance indicators of fiber optic gyroscope assembly.

[0202] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solutions of the present invention by utilizing the methods and techniques disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the protection scope of the technical solutions of the present invention.

[0203] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A method for evaluating the assembly performance of fiber optic gyroscopes based on digital twins, characterized in that... include: Establish evaluation criteria for the assembly performance of fiber optic gyroscopes and determine the corresponding quantitative indicators for each evaluation criterion. Establish a digital twin model of the fiber optic gyroscope assembly process; The assembly data of basic operators for various fiber optic gyroscope assembly performance evaluation standards were collected using digital twin models. Quantitative indicators are calculated based on the collected data to obtain the performance values ​​of each quantitative indicator. A neural network model was established and trained to obtain an evaluation model for the assembly performance of fiber optic gyroscopes. The performance values ​​of various quantitative indicators were used as input data to obtain the evaluation results of the assembly performance of fiber optic gyroscopes.

2. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 1, characterized in that: The evaluation criteria for the assembly efficiency of the fiber optic gyroscope include personnel workload, equipment level, human-computer interaction, and assembly efficiency, among which: The quantitative indicators of personnel workload are: upper limb task load, mobility task load, operational error rate, and reaction speed; The quantitative indicators of equipment level are: degree of automation, degree of digitalization, and equipment failure rate; The quantitative indicators of human-computer interaction are: logical complexity, operational complexity, and degree of logistics automation. The quantitative indicator for assembly efficiency is: production efficiency.

3. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 1, characterized in that: The digital twin model is established based on the fiber optic gyroscope assembly scenario, operator, and equipment information. It is used to collect basic operator assembly data corresponding to each quantitative indicator for quantitative indicator calculation.

4. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 2, characterized in that: The method for calculating the quantitative indicator effectiveness value of the personnel workload is as follows: Upper limb workload u 11 : In the formula, This represents the variance of the force at the elbow joint during upper limb movement. This represents the variance of the torque at the elbow joint during upper limb movement. Mobile task load u 12 : u 12 =∫0 t Mg|h t | In the formula, M is the load of the assembly worker, g is the acceleration per kilometer, and |h t | represents the absolute value of the vertical movement of the assembly worker's center of gravity; Operational error rate u 13 : In the formula, n + This represents the number of instructions correctly judged by the assembly equipment within a certain period of time, and N represents the total number of instructions issued by people within that period of time. Response sensitivity u 14 : In the formula, t1, t2, ..., t n The delay data for each event is represented by the average value, which characterizes the sensitivity of the system.

5. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 4, characterized in that: The variance of the force at the elbow joint during upper limb movement Variance of torque at the elbow joint during upper limb movement The calculation method is as follows: In the formula, F1 is the joint force on the first joint, T1 is the joint torque on the first joint, and F i T represents the joint force acting on the i-th joint. i Let F be the joint torque acting on the i-th joint, G1 and G2 be the external load weight and the weight of the forearm, respectively. g F m T g T m F and T represent the active force, passive force, active torque, and passive torque borne by the current joint drive segment, respectively. F and T are the total joint torques.

6. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 4, characterized in that: The method for calculating the effectiveness value, a quantitative indicator of the equipment level, is as follows: Automation level u 21 : In the formula, n A1 n represents the number of tasks completed automatically. A Indicates the total number of tasks; In the formula, n A2 n represents the number of processes requiring human intervention. T This indicates the total number of processes involved in the production of a single product. you 21 =0.6A1+0.4A2 Level of digitalization: In the formula, n D N represents the number of data points collected. D This indicates the theoretically collectable number of data points; Equipment failure rate u 23 : In the formula, n b N represents the number of time periods during which a failure occurred. O This indicates the total number of time periods during which the equipment operates.

7. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 4, characterized in that: The calculation method for the quantitative performance indicators of the human-computer interaction is as follows: Logical complexity u 31 : u 31 =H L In the formula, H L The entropy value of the human-computer interaction process; Operational complexity u 32 : U 32 =H O In the formula, H O The second-order entropy value in the human-computer interaction process; Logistics automation level U 33 : In the formula, n T N represents the number of transportation tasks completed automatically. T This indicates the total number of transportation tasks.

8. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 7, characterized in that: The first-order entropy value H of the human-computer interaction process L The calculation method is as follows: In the formula, n is the number of node types in PCG, p(A i ) represents the node i belonging to the Ath node. i The probability of a node of a certain type; The second entropy value H of the human-computer interaction process O The calculation method is as follows: In the process of calculating the first-order entropy value, the classification of nodes is based on the human-computer interaction process in the digital twin model. In the program control flow (PCG), the actions of humans or machines are used as nodes, and the information interaction between different actions is used as the connection to achieve node selection. During the second-order entropy calculation, the nodes are grouped according to the names of different steps in the program control process (CLP) of the digital twin model. When the names of adjacent nodes are different, the number of groups is different and the second-order entropy calculation results are different.

9. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 7, characterized in that: The calculation method for the quantitative indicator of assembly efficiency is as follows: Production efficiency u 41 : In the formula, t is the total time to complete all tasks, and n is the number of tasks.

10. The method for evaluating the assembly performance of a fiber optic gyroscope based on digital twins according to claim 7, characterized in that: The neural network model is configured with 11 neurons in the input layer, 1 neuron in the output layer, and 64 neurons in the hidden layer. Historical usage data on the evaluation of the fiber optic gyroscope assembly process is used as the training set for the neural network model. Twenty models were selected for training. 50% of the evaluation data was used as the training set, and the remaining 50% as the test set. All data was normalized to obtain the trained fiber optic gyroscope assembly performance evaluation model. The normalization method was as follows: In the formula, y i x is the normalized index value. i x represents the original index value. min and x max These represent the minimum and maximum values ​​of the quantitative indicator in the historical dataset, respectively.

Images