Intelligent tool selection method based on fuzzy ontology

Abstract

The application relates to the technical field of machining, and discloses a tool intelligent selection method based on a fuzzy ontology, which comprises the following steps: receiving macro-geometric parameters of a target gear and extracting tool nominal parameters to construct an ontology instance; combining micro-feature classification to generate a boundary compensation coefficient and a compression rate factor, performing translation and contraction calculation on a basic trapezoidal membership function model to output a reconstructed trapezoidal membership function; adjusting a model for processing tool nominal parameters by using a degree fuzzy limiting operator to generate a multi-dimensional membership vector set; judging a matching state, generating rough machining and fine machining task individuals when there is a matching conflict, issuing a fuzzy modification instruction to reset the operator strength and performing secondary logical deduction; extracting a minimum comprehensive deviation distance individual after secondary logical deduction as a result output to generate process coordination data. The application decouples the physical performance contradiction under the multivariable constraint condition, avoids logical deadlock in single tool selection and matching, and improves the matching accuracy of a process planning sequence.

Description

Technical Field

[0001] This invention relates to the field of machining technology, specifically to a method for intelligent tool selection based on fuzzy ontology. Background Technology

[0002] In the complex gear cutting process, the selection of cutting tools directly affects the final forming quality and machining feasibility. Existing tool selection methods mostly rely on trial and error based on manual experience or comparison of basic physical parameters. However, the physical states of different cutting regions of a gear differ significantly. For example, the tooth surface emphasizes micro-geometric precision, while the tooth root needs to withstand larger impact loads. Traditional single-parameter comparison methods lack objective algebraic calculation mechanisms based on feature mapping, making it difficult to handle matching conflicts under multivariate constraints in different regions during gear cutting. This results in a low degree of consistency between the final tool selection and the actual physical requirements of the process.

[0003] Meanwhile, when dealing with complex working conditions, existing tool selection logic often forces a single tool to cover the entire cutting process. Since the physical properties of tools, such as wear resistance and toughness, are subject to conflicting constraints in materials science, when the performance requirements of multiple dimensions are all at a high level, the system is very prone to deadlock in the tool selection logic due to parameter conflicts, and cannot output a tool entity that satisfies all extreme conditions, thus causing the entire process planning sequence to lose its actual executability.

[0004] Furthermore, existing tool performance evaluation mechanisms typically employ static physical tolerance models, lacking the ability to dynamically approximate time-varying cutting load characteristics. When screening candidate tool sets, existing technologies cannot adaptively adjust the boundary coordinates of the evaluation function based on actual machining characteristics, nor can they accurately quantify the comprehensive deviation between the physical properties of candidate tools and the ideal evaluation extreme value. Consequently, they cannot output reasonable process-coordinated tool combinations under strict boundary condition constraints. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a tool intelligent selection method based on fuzzy ontology, which solves the problem that in the process of complex gear cutting and forming, the conflict between the physical requirements for impact resistance in the tooth root region and the micro-geometric accuracy requirements in the tooth surface region leads to a single tool selection falling into a matching deadlock or a decline in physical performance.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a tool intelligent selection method based on fuzzy ontology, comprising the following steps: The macroscopic geometric parameters of the target gear are received, and the nominal tool parameters in the cutting standard data are combined to establish an ontology instance in the fuzzy ontology knowledge base, and the cutting region containing micro-feature classification is divided. A basic trapezoidal membership function model is established, and the numerical mapping relationship between macroscopic geometric parameters and microscopic feature classification is extracted. Boundary compensation coefficient and compression factor are calculated and generated. The coordinate boundary of the basic trapezoidal membership function model is translated and contracted using the boundary compensation coefficient and compression factor to generate a reconstructed trapezoidal membership function. Substitute the tool nominal parameters into the reconstructed trapezoidal membership function and perform algebraic mapping to obtain the initial membership degree values. Establish a degree fuzzy constraint operator adjustment model to allocate operator strength for different cutting regions. Perform base exponentiation to generate a multidimensional membership degree vector set. Determine the matching status between the multidimensional membership vector set and the set threshold conditions. Under the condition that the multidimensional performance evaluation causes the matching failure due to parameter conflict, generate multi-stage processing task individuals based on the ontology instance, issue fuzzy modification instructions to the processing task individuals to reset the operator strength, and trigger secondary logic deduction. The tool selection results are received from the logical deduction output and encapsulated into structured data, which is then sent to an external process planning system to execute specific machining.

[0007] This invention constructs a fuzzy mathematical mapping system with multi-dimensional constraints, transforming empirical trial-and-error into numerical algebraic operations. When dealing with mutually exclusive extreme operating conditions, a state circuit breaker and task decomposition mechanism is introduced to decouple contradictory constraints in the physical performance dimension. Operator strength parameters are independently allocated according to different process stages, ensuring that process collaboration data satisfying boundary conditions is output without compromising the existence of the underlying verification logic.

[0008] Preferably, the macroscopic geometric parameters consist of the module, pressure angle, and helix angle, and the nominal parameters of the cutting tool consist of wear resistance, hot hardness, machining accuracy, and toughness. The microscopic features are classified into the tooth tip region, tooth surface region, and tooth root region. The extracted tool nominal parameters are converted into normalized values ​​within a closed interval of zero to one by performing feature scaling and then written into the ontology instance.

[0009] Preferably, during the feature scaling process, the maximum and minimum values ​​of the tool nominal parameter set within the candidate range are extracted to calculate the range value, and the relationship between the range value and the set physical discrimination threshold is determined. When the range is less than the physical discrimination threshold, the normalized value is directly assigned as the unbiased median. When the range is not less than the physical discrimination threshold, the normalized value is obtained by dividing the difference between the tool nominal parameter and the minimum value by the range value, thus avoiding the calculation anomaly of the denominator approaching zero.

[0010] Preferably, for the tooth root region with time-varying load characteristics, the module and helix angle are extracted as mapping dependent variables to generate boundary compensation coefficients and compressibility factors. The boundary compensation coefficients and compressibility factors are used to perform translation and contraction calculations on the initial boundary point coordinates of the basic trapezoidal membership function model. The translation calculation uses the impact physics compensation increment generated by combining the modulus to accumulate the right boundary points of the basic trapezoidal membership function model, and uses a dual joint truncation constraint mechanism to limit the translation range; The shrinkage calculation uses the vibration attenuation compensation amount generated by combining the helix angle and the maximum allowable interval shrinkage physical threshold generated by combining the maximum compressibility factor to approximate and scale the left boundary interval of the basic trapezoidal membership function model.

[0011] Preferably, before or after performing boundary reconstruction calculation, the length of the actual input transition interval is extracted and compared with the set physical tolerance of the transition interval; If the length of the transition interval is less than the physical tolerance of the transition interval, the corresponding transition interval is downgraded to a step function output that is either zero or one.

[0012] Preferred methods for generating multidimensional membership vector sets include: Substitute the normalized values ​​into the reconstructed trapezoidal membership function to obtain the initial membership values. If there is missing data, reset the initial membership values ​​to the logical extremum constant zero. Fixed real numbers are obtained by combining the absolute values ​​of the modulus and the spiral angle with a linear interpolation mapping mechanism, and these fixed real numbers are used as the operator strengths for the corresponding micro-feature classification. The initial membership values ​​are set as the base and the operator strength is set as the exponent to generate the modified final membership values, which are then combined to form a multidimensional membership vector set.

[0013] Preferably, the specific execution logic for determining the matching status includes: The final membership degree of the multidimensional membership degree vector set is cross-compared with the multidimensional performance reception threshold. A successful match is determined when the final membership degree of all dimensions is greater than the multidimensional performance reception threshold. A matching conflict is determined when no individual tool fully satisfies the multidimensional performance acceptance threshold condition. When a matching conflict occurs, the state conflict identifier in the system memory is erased, and a processing sequence containing roughing task individuals and finishing task individuals is generated.

[0014] Preferably, the specific execution logic based on task decomposition triggering secondary logic deduction includes: Different fuzzy modification instructions are issued for roughing and finishing tasks, and the operator strength is reset independently for different processing stages. The reset operator strength is used as a new exponent to perform a power operation to obtain the secondary modification membership degree. Based on the recalculated secondary modification membership degree, the first tool subset and the second tool subset that satisfy the multidimensional performance acceptance threshold condition are extracted.

[0015] Preferably, when extracting the tool selection results, a multi-dimensional weighted Euclidean distance evaluation logic is introduced: The physical performance difference is obtained by calculating the difference between the ideal membership extreme constant and the final membership or secondary modified membership. The square of the physical performance difference is multiplied by the weight coefficient of the corresponding dimension, summed, and the square root is taken to output the comprehensive deviation distance. Extract the individual combinations with the smallest overall deviation distance within the first tool subset and the second tool subset to generate process collaboration data.

[0016] This invention provides a method for intelligent tool selection based on fuzzy ontology. It has the following advantages: 1. This invention extracts the macroscopic geometric parameters and microscopic features of the target gear to generate a reconstructed trapezoidal membership function, and substitutes it into the nominal parameters of the tool to perform a base exponentiation operation to generate a multidimensional membership vector set. This transforms the traditional empirical trial-and-error method into an objective algebraic calculation based on fuzzy ontological features, which solves the matching conflict problem under multivariate constraints in different regions during gear cutting and improves the consistency between the tool selection result and the actual process physical requirements.

[0017] 2. This invention instantiates roughing and finishing task individuals when a conflict occurs in the matching of multidimensional membership vectors, and issues different fuzzy modification instructions to reset the operator strength to perform secondary logical deduction. This decouples the contradictory constraints of a single tool in terms of physical properties such as wear resistance and toughness, avoids the selection process from falling into logical deadlock, and ensures the executability of the process planning sequence under complex working conditions.

[0018] 3. This invention utilizes boundary compensation coefficients and compression factor to perform translation and contraction calculations on the boundary coordinates of the basic membership function, and combines multidimensional weighted Euclidean distance to calculate the comprehensive deviation distance of candidate tools. This achieves dynamic approximation of the evaluation model based on time-varying load characteristics, quantifies the physical performance difference between candidate entities and ideal evaluation extreme values, and outputs a process-coordinated tool combination that meets the boundary conditions. Attached Figure Description

[0019] Figure 1 This is a flowchart of the intelligent tool selection method based on fuzzy ontology of the present invention; Figure 2 This is a flowchart illustrating the ontology definition and instantiation process of the tool and gear feature information of the present invention. Figure 3 This is a flowchart of the dynamic reconstruction process of the fuzzy membership function boundary of the present invention; Figure 4 This is a flowchart illustrating the process of refining and limiting the degree of multi-dimensional performance fuzzy information in this invention. Figure 5 A flowchart illustrating the rule reasoning and dynamic instance derivation process used to assist the business middleware of this invention. Figure 6 This is a comparison diagram of dynamic reconstruction of fuzzy membership function boundaries according to the present invention, wherein, Figure 6 (a) represents the translation of the ductile boundary of the tooth root region in this invention. Figure 6 (b) is the shrinkage of the precision boundary of the tooth surface area in this invention; Figure 7 This is a comparison diagram of the multidimensional membership distribution of candidate cutting tools for this invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See attached document Figure 1 This invention provides a method for intelligent tool selection based on fuzzy ontology, comprising the following steps: S10. The instantiation module receives the macroscopic geometric parameters of the target gear, extracts the tool nominal parameters from the cutting standard data, defines them as data attributes of the ontology instance, and stores them in the fuzzy ontology knowledge base. Simultaneously, it defines the microscopic feature classifications of the tooth tip region, tooth surface region, and tooth root region in the fuzzy ontology knowledge base. Macroscopic geometric parameters include module, pressure angle, and helix angle. Tool nominal parameters include wear resistance, hot hardness, machining accuracy, and toughness. S20: The reconstruction module calls the basic trapezoidal membership function model to obtain the module and helix angle parameters entered in S10. For the tooth root region with time-varying load characteristics, the reconstruction module generates boundary compensation coefficients and compressibility factors based on the values ​​of module and helix angle. Then, it uses the boundary compensation coefficients and compressibility factors to perform translation and contraction calculations on the initial boundary point coordinates of the basic trapezoidal membership function, and outputs the reconstructed trapezoidal membership function with topological extremum constraints. S30, the precision module extracts the nominal parameters of individual tools in the fuzzy ontology knowledge base, substitutes them into the reconstructed trapezoidal membership function generated by S20 to perform algebraic calculations and obtain the initial membership degree values, and establishes a degree fuzzy constraint operator adjustment model. According to different micro-features, the corresponding operator strengths are classified and assigned, and then the initial membership degree values ​​are subjected to exponential operations to obtain a multidimensional membership degree vector set. S40. The inference module includes a business middleware based on code logic and a semantic rule inference engine. The business middleware calls the semantic rule inference engine to read the multi-dimensional membership vector set to determine the matching status of entity data that meets the multi-dimensional performance threshold conditions. When multivariate parameters cause matching conflicts, the semantic rule inference engine writes a state conflict identifier to the part individual. The business middleware listens for the state conflict identifier to trigger runtime intervention, erases the state conflict identifier, and instantiates a roughing subtask individual and a finishing subtask individual based on the original part individual. Then, it issues different fuzzy modification instructions to the roughing subtask individual and the finishing subtask individual, and triggers the semantic rule inference engine to perform secondary logical deduction. S50: The output module receives the rule parsing feedback data from the inference module, extracts the single instance matching result, or extracts the process coordination data containing roughing high-toughness tools and finishing high-wear-resistant tools. Then, it encapsulates the process coordination data into a structured selection result format and sends the structured selection result data to the external process planning system.

[0022] See attached document Figure 2 This section describes the ontology definition and instantiation process of the tool and gear feature information in step S10. Specifically, step S10 further includes the following sub-steps: S11. In this embodiment, the instantiation module establishes basic classes and micro-feature subclasses in the fuzzy ontology knowledge base. The instantiation module uses the web ontology language to establish basic categories in the tool selection domain. The basic categories include part classes, performance condition classes, tool classes, and fuzzy set classes. For the nonlinear force distribution state in the gear forming cutting process, the instantiation module extends the gear micro-geometric feature subclass under the part class, and divides the gear micro-geometric feature subclass into tooth tip region, tooth surface region, and tooth root region. The tooth tip region has the attribute of large material removal allowance. The tooth surface region has the attributes of high relative sliding speed and severe friction. The tooth root region has the attributes of instantaneous cutting thickness abrupt change and easy tool interference. The instantiation module further subdivides the part class into tooth tip region, tooth surface region, and tooth root region. Based on the region subdivision logic, the instantiation module discretizes the macroscopic gear part into cutting load-bearing units with independent boundary conditions.

[0023] For the underlying code writing logic of building base classes and subclasses using web ontology languages, those skilled in the art can use open-source ontology editor tools to complete the development. Its syntax structure based on resource description framework is a well-known technology in this field and will not be elaborated here.

[0024] S12. The instantiation module receives the macroscopic geometric parameters of the target gear and performs a part instantiation operation. As a preferred method, the instantiation module obtains the drawing input parameters of the target gear and extracts the module, pressure angle, and helix angle as macroscopic geometric parameters based on the geometric correlation of the parameters. The module determines the absolute depth of cut and material removal rate; the pressure angle affects the curvature change of the involute profile; and the helix angle determines the magnitude of the axial force during the cutting process.

[0025] Subsequently, the instantiation module creates a specific part individual in the fuzzy ontology knowledge base to represent the gear object to be processed, and defines the acquired module, pressure angle, and helix angle as data attributes of the part individual. The data attribute definitions are persistently stored using double-precision floating-point format. The specific physical access mechanism and database construction logic for persistent data storage using double-precision floating-point format can be implemented using existing relational or non-relational databases by those skilled in the art; the underlying data read / write technology is well-known in the field and will not be elaborated upon here.

[0026] S13. The instantiation module extracts the nominal tool parameters from the cutting standard data and performs tool instantiation. The instantiation module reads the metal cutting handbook or the standard database provided by the tool manufacturer, retrieving available tool model data. For each tool model, the instantiation module creates an independent tool instance under the tool class in the fuzzy ontology knowledge base, and simultaneously extracts the nominal tool parameters. The nominal tool parameters include wear resistance, hot hardness, machining accuracy, and toughness. Wear resistance characterizes the tool's ability to resist scratching and frictional wear from hard particles. Hot hardness characterizes the tool's ability to maintain its yield strength under high-temperature cutting conditions. Machining accuracy characterizes the manufacturing tolerance level of the tool's cutting edge profile. Toughness characterizes the tool matrix's ability to resist impact loads.

[0027] In this embodiment, the instantiation module defines wear resistance, hot hardness, machining accuracy, and toughness as data attributes for each individual tool instance. Considering the impact of different physical dimensions on the subsequent logic deduction engine's calculations, the instantiation module converts the numerical values ​​of each tool's nominal parameters into normalized values ​​within the real number range of 0 to 1 using a linear mapping. The linear mapping calculation logic is based on a feature scaling formula. The feature scaling normalization formula is: ; in, This represents the normalized calculated nominal tool parameter value; This represents the initial nominal tool parameter value recorded in the database for the current tool instance. This represents the minimum value in the set of nominal tool parameters within the current candidate range; This represents the maximum value in the set of nominal tool parameters within the current candidate range.

[0028] During the execution of the feature scaling and normalization formula, to ensure the integrity of the underlying algorithm logic, the instantiation module has a built-in anti-lockdown mechanism for exceptions where the denominator approaches zero. Specifically, the instantiation module sets the value range to 1×10. -6 Up to 1×10 -5 Physical discrimination threshold The physical discrimination threshold is set based on the truncation error tolerance during computer double-precision floating-point calculations. The instantiation module executes the absolute value of the deviation conditional judgment logic: when... If the condition is true, it means that the nominal parameters of the corresponding categories of all tools in the current candidate library do not have physical distinguishability. In this case, the instantiation module skips the division operation in the feature scaling normalization formula and directly applies the normalized nominal parameter values. We assign 0.5 as the unbiased median. The unbiased median ensures that the corresponding tool nominal parameters do not have a biased impact on decisions in any dimension, thus completely avoiding the risk of algorithm execution crash caused by a denominator of zero.

[0029] S14. The instantiation module establishes object attribute associations between micro-feature subclasses and tool nominal parameters. After completing individual instantiation and data attribute assignment, the instantiation module defines object attribute connection relationships in the fuzzy ontology knowledge base. Combining the specific working conditions of the micro-region, the instantiation module establishes performance requirement association mappings for different machining areas. Specifically, it establishes object attribute connections between the tooth surface area and high wear resistance and high machining accuracy, and establishes object attribute connections between the tooth root area with time-varying load characteristics and high toughness. The establishment of object attribute associations establishes the underlying data retrieval path.

[0030] See attached document Figure 3 This section describes the dynamic reconstruction process of the fuzzy membership function boundary based on topological security constraints in step S20. Specifically, step S20 further includes the following sub-steps: S21. The reconstruction module calls the basic trapezoidal membership function model. The basic trapezoidal membership function model is used to map discrete tool physical performance parameters to continuous fuzzy evaluation indices. The reconstruction module initializes the trapezoidal piecewise mathematical model in computer memory to build the computational foundation. The mathematical expression of the basic trapezoidal membership function is: ; in, Represents the basic trapezoidal membership function; The basic physical performance variables representing the input are the nominal tool parameter values ​​calculated after normalization in step S10. ; The left absolute zero boundary point represents the basic trapezoidal membership function; The left-hand complete membership boundary point represents the basic trapezoidal membership function; The right-hand boundary point of the basic trapezoidal membership function represents the complete membership boundary point. The right-hand absolute zero boundary point represents the membership function of the basic trapezoid; , , and Together they form the initial boundary point coordinates of the basic trapezoidal membership function; Represents a logical extremum that is completely unaffiliated; Represents the logical extremum of complete membership; The left-side performance transition rate represents the left-side performance transition rate when the tool's nominal parameters are in the left-side transition range. The right-side performance transition rate represents the performance transition rate when the nominal parameters of the cutting tool are in the right-side transition range. The logical judgment boundary representing the nominal parameters of the tool falling into the physical failure range on the left; The logical judgment boundary represents the nominal parameters of the tool being in a performance ramp-up state on the left. The core logical judgment boundary represents the nominal parameters of the cutting tool satisfying the optimal physical matching state. The logical threshold representing the tool's nominal parameters being in a state of performance degradation on the right side; This represents the logical threshold for determining whether the tool's nominal parameters fall within the physical failure range on the right. The refactoring module establishes independent execution instances for wear resistance, hot hardness, machining accuracy, and toughness based on the fundamental trapezoidal membership function model.

[0031] In this embodiment, to ensure the integrity of the underlying operations and completely avoid arithmetic overflow crashes caused by the denominator approaching zero during division, the reconstruction module incorporates denominator extreme value verification logic before performing the calculation. Specifically, this is achieved by setting the value range to 1×102. -6 Physical tolerance of the transition zone Physical tolerance of the transition zone Based on the truncation error safety tolerance of the underlying architecture of computer IEEE 754 double-precision floating-point numbers, it is obtained by calibrating the minimum resolvable machine precision of the hardware processor.

[0032] The denominator extremum verification logic, as a built-in property of the function model, takes effect for any input actual evaluation interval. That is, regardless of whether it's for the basic boundary parameters or the subsequently reconstructed boundary parameters, when determining the length of the left transition interval of the actual input (e.g., ... Or reconstructed ) When the parameter is less than or equal to the critical threshold of the corresponding interval, the reconstruction module will downgrade the corresponding left transition interval to a step function. That is, when the parameter is less than or equal to the critical threshold of the corresponding interval, the function value will be directly output as 0, and when it is greater than the critical threshold, the function value will be directly output as 1. Similarly, when determining the length of the right transition interval of the actual input... At that time, the refactoring module performs an equivalent step degradation process on the right-hand transition interval. For the logic of writing conditional branch statements to construct piecewise function logic in the underlying computer architecture, those skilled in the art can use existing standard programming language syntax libraries. The conditional judgment execution flow technique is a well-known technique in this field and will not be elaborated upon here.

[0033] S22. As a preferred method, the reconstruction module extracts key physical parameters as input dependent variables based on the principle of metal cutting dynamics, thereby obtaining the modulus entered in step S10. With helix angle To generate boundary compensation coefficients and compression ratio factors. During gear cutting and forming, the cutting layer thickness changes periodically with the tool rotation position. Therefore, for the tooth root region with time-varying load characteristics, the reconstruction module extracts parameters in response to the nonlinear impact of the underlying physical machining environment. Module selection. and helix angle As an input parameter, it is because of: modulus The spatial span of the tooth groove and the volume of metal removed are determined by the module. It shows a clear positive correlation with the instantaneous impact force during the cutting process; helix angle It determines the axial deflection component when the tool enters the workpiece and the amplitude of alternating vibration during the cutting process.

[0034] The reconstruction module retrieves preset cutting physical constants from the fuzzy ontology knowledge base to establish reconstruction operation factors. The cutting physical constants include the first rigidity compensation coefficient. With the second rigid compensation coefficient First rigid compensation coefficient With the second rigid compensation coefficient These lower-level characteristic parameters together constitute the boundary compensation coefficients. First rigid compensation coefficient. The effective value range is (0.01, 0.05), the second rigid compensation coefficient. The effective value range is (0.05, 0.15). First rigid compensation coefficient. With the second rigid compensation coefficient The value is determined based on the dynamic load data of the three-dimensional force measuring instrument in historical cutting force test experiments, combined with the mean filtering of multiple sets of orthogonal cutting experiments, and obtained by stress curve fitting and calibration using the least squares method.

[0035] To ensure fitting accuracy, the reconstruction module incorporates multi-source data time-domain alignment logic before executing the least squares method. Based on high-frequency microsecond-level timestamps, it performs hard synchronization matching between the load fluctuation sequence output by the three-axis force gauge and the phase sequence output by the spindle encoder, thereby eliminating phase differences caused by nonlinear measurement and control delays. For the specific mathematical derivation of stress fitting curve calibration using the least squares method, those skilled in the art can refer to standard numerical analysis methods; the data fitting algorithm is well-known in the field and will not be elaborated upon here. The reconstruction module simultaneously retrieves the maximum compression factor. Maximum compressibility factor The limit of shrinkage deformation is a real number with a fixed range and is used to limit the range of operations. It is determined by the maximum allowable strain safety threshold derived from the matrix material before plastic yielding.

[0036] S23. Based on the principle of boundary constraint topological deformation, after constructing the underlying mapping architecture according to the principle of yield limit strain compensation in macroscopic mechanics of materials, the reconstruction module uses the boundary compensation coefficient and compressibility factor to perform translation calculations on the initial boundary point coordinates of the basic trapezoidal membership function. Specifically, the reconstruction module executes the first set of mathematical coordinate transformation logic to output the right-side reconstruction point with topological extremum constraints, thereby performing translational expansion compensation to meet the high toughness requirements of the tooth root interference zone. That is, as the module... The reconstruction module adds a feature whereby it shifts the fuzzy set boundary points representing high toughness requirements to the right, simultaneously calculating the rightward drift of both the right-side fully subordinate boundary points and the right-side absolute zero boundary points. Furthermore, to prevent coordinate points from exceeding the effective range of the real number domain, while also considering the mechanical fatigue limit of the tool matrix material, the reconstruction module introduces a minimum-value function to establish a dual joint truncation constraint mechanism based on the upper bound dimension of the mathematical domain and the material physical safety truncation dimension. The formula for the rightward shift is: ; ; in, This represents the right side of the reconstructed boundary point; This represents the right-hand absolute zero boundary point after reconstruction; The right-hand boundary point of the basic trapezoidal membership function represents the complete membership boundary point. The right-hand absolute zero boundary point represents the membership function of the basic trapezoid; Represents the first rigid compensation coefficient; Represents the modulus; Represents modulus-based The generated impact physics compensation increment; A function that takes the minimum value in a set; The right-hand side fully belongs to the boundary theory translation point without extreme value constraints; The theoretical translation point representing the right-hand absolute zero boundary without extreme value constraints; The upper limit extreme value represents the normalized domain interval; This represents the set performance resolution, which in this embodiment is set to 1×10. -4 ; This represents the upper limit of the truncation constraint approaching a safe value. Performance resolution. The minimum resolvable performance difference gradient of the tool matrix material in microscopic physical metallographic testing is determined to ensure that the right-side fully subordinate boundary point after reconstruction is less than the right-side absolute zero boundary point after reconstruction, thereby maintaining the mathematical existence of the right-side slope of the trapezoidal function.

[0037] S24. The reconstruction module uses boundary compensation coefficients and compression factors to perform shrinkage calculations on the initial boundary point coordinates of the basic trapezoidal membership function. Before performing the centripetal approximation shrinkage calculation, it establishes a mathematical correlation mapping based on the forced vibration attenuation tolerance principle in mechanical dynamics. On this basis, the reconstruction module executes the second set of mathematical coordinate transformation logic to output the left reconstruction point with topological extremum constraints, and then performs centripetal shrinkage to meet machining accuracy requirements. Since alternating vibration caused by axial force reduces actual cutting accuracy, the helix angle... The reconstruction module adds a centripetal approximation contraction to the left-side performance transition range to meet high machining accuracy requirements. Simultaneously, to prevent the left-side absolute zero boundary point from exceeding the left-side fully subordinate boundary point, thus causing cross-flipping errors in the function graph, the reconstruction module incorporates the maximum compression factor. Execute the anti-lock braking mechanism. The formula for calculating centripetal contraction is: ; ; in, This represents the reconstructed left absolute zero boundary point; This represents the left side, which is completely subordinate to the boundary point after reconstruction. The left absolute zero boundary point represents the basic trapezoidal membership function; The left-hand complete membership boundary point represents the basic trapezoidal membership function; This represents the second rigid compensation coefficient; Represents the helix angle; Represents the sine function; It represents the absolute value of the sine function calculation result, used to avoid abnormal negative oscillation components caused by the reverse input of the spiral angle; Represented by helix angle The generated vibration attenuation compensation amount; Represents the maximum compression factor; The initial length of the left transition interval represents the basic trapezoidal membership function; Represented by the maximum compression ratio factor The physical threshold for the maximum permissible range of contraction. Maximum compressibility factor. The effective range of values ​​is constrained to the interval (0, 0.45). The physical meaning of the constraint logic is that, since the left and right endpoints of the interval each shrink to the center by a maximum of 0.45 times the original interval length, the total maximum shrinkage is 0.9 times the original interval length, ensuring that at least 10% interval margin is always maintained, thereby preventing interval flipping anomalies from a mathematical perspective. It represents a function that takes the minimum value in a set.

[0038] The reconstruction module replaces the corresponding coordinate parameters in the original basic trapezoidal membership function with the reconstructed left absolute zero boundary point, the reconstructed left fully membership boundary point, the reconstructed right fully membership boundary point, and the reconstructed right absolute zero boundary point. Finally, the reconstruction module outputs a reconstructed trapezoidal membership function with time-varying adaptive characteristics, completing the dynamic topological restructuring of the boundary point coordinates. For the data mapping update operation of writing the reconstructed boundary point coordinates into the memory structure to replace the corresponding coordinate parameters of the original function, those skilled in the art can use the attribute overloading mechanism in current object-oriented programming. Its memory address addressing and data overwriting techniques are well-known in the field and will not be elaborated upon here.

[0039] See attached document Figure 4 This section describes the process of refining and limiting the degree of multi-dimensional performance fuzzy information in step S30. Specifically, step S30 further includes the following sub-steps: S31. In this embodiment, the precisionization module extracts the normalized nominal tool parameter values ​​generated in step S10 from the fuzzy ontology knowledge base. The initial membership values ​​are then substituted into the reconstructed trapezoidal membership function generated in step S20 to perform algebraic calculations, and subsequently, floating-point mapping is performed in computer memory to obtain the initial membership values. The initial membership values ​​are distributed within the closed real number interval from 0 to 1. To support subsequent vector operations, the refinement module establishes a multi-dimensional storage matrix, writing the initial membership values ​​corresponding to the wear resistance data attribute, thermohardness data attribute, machining accuracy data attribute, and toughness data attribute as independent elements into the multi-dimensional storage matrix.

[0040] Before performing the matrix write operation, the precision module introduces data integrity verification logic. Specifically, the precision module checks whether all dimensional attributes of the current tool individual have valid floating-point numbers. If missing values ​​or non-numeric identifiers are detected, the precision module forcibly assigns the corresponding initial membership value to the logical extremum constant zero, thereby completely avoiding underlying computational anomalies caused by null pointers or dimension mismatches in the underlying matrix operations. Given that zero represents a completely non-membered physical failure state in the underlying trapezoidal membership function model, setting the logical extremum constant zero as a penalty parameter threshold ensures that tool individuals with missing data are directly eliminated in subsequent multi-dimensional collaborative reasoning, thus guaranteeing the absolute safety of process selection. For the allocation and element reading / writing logic of multi-dimensional array matrices in computer memory, those skilled in the art can use standard data structure libraries. The memory address allocation and contiguous addressing techniques are well-known in the field and will not be elaborated upon here.

[0041] S32. The refinement module establishes a degree fuzzy constraint operator adjustment model. As a preferred approach, before performing specific algebraic modification calculations, the refinement module constructs mathematical associations according to the semantic polarity strengthening and weakening principles in fuzzy mathematics. In fuzzy logic theory, conventional membership degree calculation results cannot characterize the differences in the intensity of requirements with semantic polarity, such as extremely stringent or slightly lenient requirements. To reflect the stringency requirements of micro-feature classification on cutting performance, the refinement module introduces a degree fuzzy constraint mechanism based on exponential operations. Based on the underlying mapping architecture, the refinement module performs exponential operations on the initial membership degree values ​​to achieve degree modification, thereby obtaining a multi-dimensional membership degree vector set. The calculation formula is as follows: ; in, This represents the final membership degree after modification; The basic physical performance variables representing the input are the nominal tool parameter values ​​after normalization calculation in step S10. This means substituting the normalized tool nominal parameter values ​​into the initial membership values ​​calculated by reconstructing the trapezoidal membership function; Represents operator strength; This represents the result of an exponential operation on the initial membership values, with the initial membership values ​​as the base and the operator strength as the exponent.

[0042] To ensure the rigor of underlying operations and prevent undefined mathematical exceptions caused by a zero base and a non-positive exponent, the precision module constrains the operator strength before performing exponentiation operations. The value of is always greater than zero. If the program detects an abnormal external input causing a decrease in operator strength... For non-positive, precise modules will force the injection of a minimum threshold of 1×10. -3To maintain the existence of mathematical functions, exponential operations change the distribution of initial membership values ​​through nonlinear mapping, thereby tightening or loosening the underlying physical constraints. The aim is to transform the implicit evaluation experience of cutting experts into numerical penalty or reward weights that can be precisely executed by computers.

[0043] S33. The refinement module classifies and assigns corresponding operator strengths based on different micro-features, and formulates operator strength allocation rules based on the load-bearing characteristics of the solid cutting area. To ensure the completeness of the underlying data for multi-dimensional matrix operations and prevent logical omissions, the refinement module has a built-in default assignment mechanism. For common data attributes that do not form a strong correlation with the current specific micro-feature classification (such as thermal hardness, which is not extremely required in this embodiment), the refinement module uniformly assigns a constant constant of 1.0 as the baseline operator strength. When the operator strength is 1.0, the exponential operation degenerates into an identity mapping, that is, the final membership degree of the corresponding dimension is equal to the initial membership degree, thus representing the maintenance of the original physical evaluation weights.

[0044] For the tooth root region, which experiences sudden changes in cutting thickness and is prone to tool interference, the precision module applies emphasized constraint logic. Because the tooth root region places extremely stringent demands on the impact resistance of the tool matrix, and insufficient impact resistance can easily lead to chipping failure, the relevant underlying calculations prioritize the application of emphasized constraint logic. The precision module assigns fixed real numbers greater than 1 as operator strengths to the toughness data attributes associated with the tooth root region.

[0045] In this embodiment, the operator strength value assigned to the toughness data attribute is defined as the interval (1.5, 3.0). To enable the algorithm to obtain a unique calculation factor from a continuous interval, the precision module incorporates a linear interpolation mapping mechanism. Specifically, the module of the target gear is read as the mapping benchmark. Based on the positive correlation between module and impact load, when the module is at the minimum value of the set of parts to be processed, the lower limit of the interval, 1.5, is extracted as the deterministic operator strength; when the module reaches the maximum value of the set of parts to be processed, the upper limit of the interval, 3.0, is extracted as the deterministic operator strength; the module in the intermediate state is derived as a unique fixed real number within the interval (1.5, 3.0) through linear interpolation. The establishment of this interval is based on the nonlinear regression analysis results of tool fracture failure rate and toughness data attribute in historical cutting life experiments to ensure the effective identification of critical performance individuals.

[0046] For constructing nonlinear regression analysis models for historical experimental data, those skilled in the art can use the least squares method combined with polynomial curve fitting. The underlying regression analysis fitting calculations are well-known techniques in this field and will not be elaborated upon here. Mathematically, when the operator strength is greater than 1 and the base is within the range of 0 to 1, the exponential operation causes the initial membership value to exhibit a nonlinear decay trend. This effectively reduces the final score of individual tools that are at the edge of their physical performance, thereby filtering out tool models with critical impact resistance in the logical screening stage.

[0047] S34. For tooth surface regions with high relative sliding speed and severe friction, the precision module also matches the emphasis-type constraint logic and assigns independent fixed real numbers greater than 1 as operator strengths for the wear resistance data attribute and machining accuracy data attribute associated with the tooth surface region. As a preferred approach, the operator strength value assigned to the wear resistance data attribute is calibrated to the interval (1.2, 2.5), and the operator strength value assigned to the machining accuracy data attribute is calibrated to the interval (1.1, 2.0). Combined with the mapping value logic based on specific working conditions, the absolute value of the helix angle of the target gear is read as the mapping dependent variable, and a linear interpolation mapping mechanism similar to that of the module is used to calculate and generate a unique specific operator strength assignment within the corresponding interval. The corresponding parameter values ​​are obtained by collecting tooth surface wear mapping data and surface roughness detection data for correlation clustering calibration.

[0048] For the correlation clustering and labeling operation of multidimensional mapping data, those skilled in the art can use the classic K-means clustering or density-based spatial clustering algorithm. The unsupervised learning classification and cluster center calculation are well-known techniques in the field and will not be elaborated here.

[0049] Subsequently, the refinement module gathers all the final modified membership degrees and constructs attribute vector combinations with individual tooling units as basic objects in the fuzzy ontology knowledge base, thereby generating a complete multidimensional membership vector set. Each numerical element in the multidimensional membership vector set represents the comprehensive adaptability and matching degree of a single tooling model under the combined effects of specific modulus and helix angle variables and specific micro-feature classifications. Based on the global selection logic of multidimensional nonlinear constraints, the multidimensional membership vector set avoids the one-sidedness caused by relying on a single physical extreme value for decision-making. The refinement module saves the multidimensional membership vector set to persistent storage space for subsequent semantic reasoning engine to perform condition judgment and state circuit breaking inference. For the underlying logic of serializing the multidimensional data structure and saving it to persistent storage space, those skilled in the art can use existing relational databases or ontology file read / write interfaces to implement it. The data persistence and data writing technology is well-known in the field and will not be elaborated here.

[0050] See attached document Figure 5 This section describes the rule reasoning and dynamic instance derivation process assisted by the business middleware in step S40. Specifically, step S40 further includes the following sub-steps: S41. In this embodiment, the inference module includes a business middleware based on code logic and a semantic rule inference engine. Based on the need for rigorous judgment of the underlying logic, the inference module establishes multi-dimensional performance threshold conditions for entity data. These multi-dimensional performance threshold conditions use a unified multi-dimensional performance receiving threshold. (In this embodiment, the threshold is set to [0.60, 0.85]) for definition. Multidimensional performance reception threshold. The value is set based on the extreme value of the acceptable abnormal tool failure probability in historical cutting data. That is, the business middleware retrieves the statistical samples of tool failures that conform to the Weibull distribution in historical cutting life, extracts the lower boundary of the confidence interval with a cumulative failure rate of less than 5% as the calculation benchmark, and then maps it to obtain the quantitative boundary for dividing absolute safety and potential risk.

[0051] The business middleware calls the semantic rule reasoning engine to read the multidimensional membership vector set generated in step S30 and compares the numerical elements therein with... Cross-comparison is performed to determine the matching status of entity data that meets the multidimensional performance threshold conditions. The matching status is characterized by simultaneously satisfying constraints in multiple dimensions, specifically requiring that the final membership degrees of toughness, wear resistance, hot hardness, and machining accuracy of the corresponding tool individual are all greater than the multidimensional performance acceptance threshold. .

[0052] S42. When the modulus and helix angle values ​​entered in step S10 are too large, triggering extreme reconstruction and modification of the underlying membership functions, resulting in no tool individual in the multidimensional membership vector set being able to simultaneously meet the multidimensional performance threshold conditions, the semantic rule reasoning engine determines that a matching conflict has occurred. Given that the underlying description logic of the web ontology language is based on the open-world assumption and monotonic reasoning mechanism, it typically does not support non-monotonic reasoning and therefore cannot directly handle parameter deadlock and state backtracking operations. To overcome the limitation of parameter deadlock by the underlying logic, the semantic rule reasoning engine writes a state conflict flag, persistently stored using Boolean data attributes, to the original part individuals in the fuzzy ontology knowledge base. When the state conflict flag is assigned a true value, it indicates that the current reasoning process is stuck in a single tool selection deadlock state, meaning that the physical performance of any existing single tool cannot simultaneously meet the extreme metal removal rate requirements and tooth surface finish requirements.

[0053] S43. The business middleware uses an event-driven listener via an application programming interface to continuously monitor changes in the data attributes of individual parts in the fuzzy ontology knowledge base, thereby avoiding memory consumption caused by polling mechanisms and ensuring real-time state synchronization. When a state conflict flag is detected as true, the business middleware triggers runtime intervention, suspending the current single-tool inference process and changing the state conflict flag to false (i.e., erasing the state conflict flag), thus preventing the underlying rule engine from getting stuck in an infinite query deadlock.

[0054] Subsequently, the business middleware, based on the original part individuals, instantiates roughing and finishing task individuals with temporal dependencies in the fuzzy ontology knowledge base (i.e., finishing must be performed after roughing). Instantiating these task individuals decomposes the original single-step cutting task into a relay machining sequence, thereby decoupling the mutually exclusive constraints between extreme impact resistance requirements and extreme surface quality requirements at the physical level. For the underlying code logic for reading and writing ontology files and monitoring their status using application programming interfaces (APIs), those skilled in the art can implement it using existing web ontology language application programming interfaces (APIs). The ontology model operation and event monitoring techniques are well-known in the field and will not be elaborated upon here.

[0055] S44. The business middleware issues different fuzzy modification instructions to the roughing and finishing task individuals to reset the operator strength allocation rules in step S30. Based on the step-by-step removal principle in metal cutting, the core physical task of the roughing stage is to withstand alternating impacts and maximize the metal removal volume. Therefore, for the roughing task individuals, the business middleware issues fuzzy modification instructions that weaken precision and strengthen toughness, resetting the operator strengths related to machining precision and toughness to precision-weakening operators respectively. (Values ​​range from (0.2, 0.5)) and toughness strengthening operator (The value range is (1.5, 3.0)).

[0056] Correspondingly, the core physical task of the finishing stage is to eliminate residual surface stress and ensure the microscopic geometric accuracy of the final tooth profile. Therefore, for each individual finishing task, the business middleware issues fuzzy modification instructions to weaken toughness and strengthen wear resistance, resetting the operator strengths related to toughness and wear resistance to toughness-weakening operators respectively. (Value range is (0.2, 0.5)) and wear-resistant strengthening operator (The value range is (1.5, 3.0)). The values ​​of the four parameters are determined by the preference weights of form and position tolerances and tool breakage risk at different cutting stages. They are obtained by nonlinear fitting calibration of cutting force and surface roughness results from multiple sets of Taguchi orthogonal tests in the expert experience database. This establishes the boundaries of the threshold ranges for each parameter to ensure the objective physical validity and feasibility of parameter resetting.

[0057] S45. The business middleware uses the reset operator strength to trigger the semantic rule reasoning engine to perform secondary logical deduction. The semantic rule reasoning engine re-executes the matching state judgment logic for roughing task individuals and finishing task individuals respectively, and then sequentially extracts the first tool subset and the second tool subset that meet the reset multi-dimensional performance threshold conditions. By issuing different fuzzy modification instructions, the secondary logical deduction guides the semantic rule reasoning engine to output tool individuals with high impact resistance in the first tool subset, and to output tool individuals that meet the tooth profile forming accuracy and surface quality in the second tool subset, thereby dynamically resolving the multi-variable conflict deadlock state by combining task splitting.

[0058] To avoid the bias in selection results caused by relying on a single physical extreme value, the inference module incorporates multi-dimensional weighted Euclidean distance evaluation logic when outputting the final tool individual. For candidates in the first and second tool subsets, the inference module extracts their final membership degree across all corresponding dimensions, substitutes it into the multi-dimensional weighted Euclidean distance formula to calculate the comprehensive deviation distance, and then extracts the individual with the smallest comprehensive deviation distance as the final physical output. The multi-dimensional weighted Euclidean distance formula is: ; in, This represents the calculated overall deviation distance; This represents the total number of evaluation dimensions in this embodiment. The value is fixed at 4 to correspond to the four nominal parameters extracted in step S10; A mathematical operator that sums over all evaluation dimensions; Representing the The weight coefficients are assigned to each evaluation dimension, and the sum of the weight coefficients of all dimensions is always equal to 1. In this embodiment, the weight coefficients are... The middleware automatically allocates features based on a pre-configured dictionary. By default, an average distribution strategy is used to ensure unbiased feature evaluation (i.e., the weights of each dimension are fixed). (corresponding to a value of 0.25); when the external process planning system issues a specific processing preference instruction, the corresponding unequal weight is retrieved from the expert experience base according to the instruction mapping for overriding; The representative of individual knives in the first The final membership degree under each evaluation dimension; The extreme value constant of the ideal membership degree in an absolutely perfect state; The physical property difference between the actual final membership degree and the ideal membership degree extreme constant; The square operation, representing the difference in physical performance, is used to nonlinearly amplify the penalty effect of severely localized bottleneck dimensions in the underlying computing power.

[0059] For the logic of integrating a semantic rule reasoning engine into a computer system and executing rule assertions and parsing, those skilled in the art can use existing semantic web rule language reasoning libraries. The rule engine parsing technology is a well-known technology in this field and will not be elaborated here.

[0060] This section describes the structured output and process application of the collaborative matching scheme in step S50. Specifically, step S50 further includes the following sub-steps: S51. In this embodiment, the output module receives rule parsing feedback data from the inference module. To ensure the underlying security of data reading and avoid memory addressing anomalies, the output module embeds empty set verification logic before performing the read operation. Specifically, the program probes the header pointer of the data packet in the rule parsing feedback data. If it is determined to be a null pointer or an illegal memory address, a system alarm is triggered and an abnormal interruption command to cut off the downstream link is output to request manual intervention, thereby ensuring the physical security of the manufacturing line.

[0061] Based on the feedback result of the empty set verification, the output module reads the result according to the state conflict flag set by the inference module during step S40. In the case where no state conflict occurs and the inference module completes a single logical deduction, the output module extracts a single instance matching result. The single instance matching result represents the tool type whose underlying physical properties meet the multi-dimensional performance acceptance threshold of the target gear cutting condition and can independently complete the target gear machining. In the case where a state conflict flag is generated and the inference module completes a second logical deduction, the output module extracts process coordination data including roughing high-toughness tools and finishing high-wear-resistant tools. The process coordination data consists of the individual with the smallest overall deviation distance within the first tool subset extracted in step S40 and the individual with the smallest overall deviation distance within the second tool subset.

[0062] For the detection of the validity of memory pointers and the protection against out-of-bounds access, those skilled in the art can use the existing operating system's underlying memory management application interface. The pointer address verification technology is a well-known technology in this field and will not be elaborated here.

[0063] S52. As a preferred approach, the output module encapsulates the process collaboration data into a structured selection result format. To establish a cross-system data interaction channel, the output module performs a data mapping and parsing operation. Considering that the entity data originates from multiple heterogeneous nodes such as the tool basic physical database and the cutting process parameter library, the output module performs multi-source data version alignment logic before encapsulation. Based on the absolute time base of the system clock, the output module extracts the system timestamp and version checksum of each data slice and constructs a multi-source data synchronization check formula. Only when the result of the multi-source data synchronization check formula is true is the serialization process allowed, thereby eliminating the risk of process data mismatch caused by asynchronous database updates. The multi-source data synchronization check formula is: ; in, A Boolean value representing the result of multi-source data synchronization verification; This represents the associated baseline version checksum extracted from the tool's basic physical database; This represents the associated baseline version check code extracted from the cutting process parameter library; This represents a logical equation that determines whether the two types of data belong to the same global process baseline version, in order to avoid matching incompatible older isolated data. This represents a logical AND operator, which outputs true only if both left and right conditions are met simultaneously. The last update timestamp representing the tool node data; The last update timestamp representing the process node data; The absolute value operation representing the difference between the two timestamps; This represents the set synchronization tolerance, combined with the maximum communication latency limit of the enterprise-level data bus. It is set to a fixed constant of 500 milliseconds. The value of this constant is obtained by performing normal distribution statistics on cross-node transmission delay data in historical communication logs and extracting the upper limit of the 99.9% confidence interval. This represents a logical inequality that indicates the time deviation is within a safe tolerance range.

[0064] Subsequently, the output module converts the OWL object data in the underlying fuzzy ontology knowledge base into JSON format that can be directly recognized by external systems. The output module defines tool basic information nodes, process parameter nodes, and assembly position nodes in the structured selection result file. The tool basic information node contains tool code, material grade, and geometric dimension data extracted from the tool basic physical database, used to guide the absolute spatial positioning of the physical tool in the machine tool coordinate system. The process parameter node contains matching cutting speed, feed rate, and depth of cut data, used to constrain the spindle output power and feed axis motor speed of the CNC machine tool. The assembly position node contains tool holder interface type and spindle tool position number data generated according to the spindle specifications, used to support the precise grasping action of the automatic tool changer.

[0065] For the underlying code implementation of converting in-memory instance objects into JSON format serialized text, those skilled in the art can use existing data serialization packages for object mapping and parsing. The data structure encapsulation technology is a well-known technology in the field and will not be elaborated here.

[0066] S53. The output module sends structured configuration result data to an external process planning system. The external process planning system includes the enterprise's internally deployed computer-aided manufacturing system and enterprise resource planning system. The output module establishes a network communication connection based on the transmission control protocol and pushes the encapsulated structured configuration result file to the computer-aided manufacturing system via the application programming interface.

[0067] To verify the integrity of network communication logic and prevent system deadlock caused by packet loss, the output module has a built-in timeout retransmission mechanism. When data is sent and no acknowledgment frame is received from the external system within a preset network response time threshold (specifically 2000 milliseconds, this parameter is based on statistical analysis of the network transmission delay under full load conditions of the historical workshop LAN and the comprehensive interaction time of the external system application layer processing response, and is obtained by extracting the 95th percentile of the distribution curve), the output module suspends the current process and re-initiates transmission after a set backoff delay time (specifically 1000 milliseconds, this parameter is based on the truncated binary exponential backoff principle to avoid instantaneous network congestion peaks, and is set to half of the network response time threshold). If all transmission attempts fail after reaching the maximum retransmission count threshold (specifically set to 3 times, obtained according to the industry standard anti-dead-loop specification of the transmission control protocol), a link circuit breaker is executed and a local error log is generated.

[0068] Upon successful transmission, the computer-aided manufacturing system parses the structured selection result data to obtain the tool's three-dimensional contour parameters and generate the corresponding CNC machining toolpath code. Simultaneously, the output module independently parses the tool code from the process collaboration data and sends it to the material management terminal of the enterprise resource planning system. Upon receiving the tool code, the material management terminal triggers a verification process in the underlying inventory database to locate the corresponding physical tool and generate a material preparation list for the workshop.

[0069] For the network communication process of establishing network socket connections between different software systems and transmitting data using application programming interfaces, those skilled in the art can implement it by calling existing standard network protocol stacks. The cross-system data interaction technology is a well-known technology in this field and will not be described in detail here.

[0070] See attached document Figure 6 and attached Figure 7 To aid in understanding the present invention, a specific application embodiment is provided, taking into account the machining conditions of large-module heavy-duty gears, which specifically includes the following steps: The instantiation module receives the drawing input parameters of the heavy-duty transmission gear and extracts the macroscopic geometric parameters, specifically including the module. Pressure angle and helix angle The instantiation module creates an individual part representing the heavy-duty transmission gear in the fuzzy ontology knowledge base, assigning its module, pressure angle, and helix angle as data attributes. Simultaneously, the instantiation module retrieves three candidate tool model data from the tool basic physics database, creating instance objects for candidate tool A, candidate tool B, and candidate tool C respectively. Candidate tool A uses coated cemented carbide, candidate tool B uses conventional cemented carbide, and candidate tool C uses cobalt-containing high-speed steel. The instantiation module extracts wear resistance, hot hardness, machining accuracy, and toughness data for the three tools, and uses a feature scaling normalization formula to obtain the normalized nominal parameter values ​​of the tools. Candidate tool A possesses high wear resistance and low toughness; candidate tool B possesses balanced basic properties; candidate tool C possesses high toughness and low machining accuracy.

[0071] The reconstruction module calls the basic trapezoidal membership function model to extract the defined basic boundary points. Taking the resilience dimension as an example, the initial boundary point coordinates of the basic trapezoidal membership function are the left absolute zero boundary point. The left side is completely subordinate to the boundary point. The right side is completely subordinate to the boundary point. and the absolute zero boundary point on the right Combining modulus With the first rigidity compensation coefficient The reconstruction module calculates the impact physical compensation increment for the tooth root region. Substitute into the formula for rightward translation: ; Obtained through calculation , .

[0072] Combination Figure 6 (a) As can be seen from the translation of the toughness boundary of the tooth root region in this invention, the toughness boundary of the tooth root region shifts to the right. This is based on the normalized calculation of the tool's nominal parameter values. The x-axis represents the initial membership degree value. In the dynamic reconstruction comparison view of the fuzzy membership function boundary constructed with the vertical axis, the basic trapezoidal membership function represented by the solid line expands the distance of the fully membership boundary point on the right side under the modulus impact compensation, and transforms into the reconstructed trapezoidal membership function represented by the dashed line, which reflects the topological deformation constraint for high toughness requirements in physical form.

[0073] For the machining accuracy dimension of the tooth surface area, combined with the helix angle Second rigid compensation coefficient and maximum compression factor Calculate the vibration attenuation compensation amount Based on the left absolute zero boundary point Completely subordinate to the boundary point on the left Substituting into the centripetal approximation contraction calculation formula: ; Obtained through calculation , .

[0074] Combination Figure 6 (b) As can be seen from the shrinkage of the precision boundary of the tooth surface region in this invention, the precision boundary of the tooth surface region shrinks. For the horizontal axis, which is the normalized calculated nominal parameter value of the tool, The vertical axis represents the initial membership degree value. The evaluation system, represented by solid lines, transforms the left transition interval of the basic trapezoidal membership function into a centripetal approximation contraction shape represented by dashed lines under the effect of helical angle vibration attenuation compensation. This forces the underlying model to shrink the transition tolerance range of machining accuracy. The reconstruction module thus outputs a reconstructed trapezoidal membership function with time-varying adaptive characteristics.

[0075] The precision module substitutes the normalized nominal parameter values ​​of each tool into the reconstructed trapezoidal membership function to perform calculations and obtain initial membership values. Based on the cutting conditions of the heavy-duty gear, the precision module assigns operator strengths. Operator strengths are also assigned based on the toughness associated with the tooth root region. Strength of wear resistance distribution operator associated with tooth surface region Substitute into the calculation formula: The process performs exponential operations to ultimately obtain a multidimensional membership vector set containing all individual tools. The numerical results of each candidate tool in different dimensions are reflected in the multidimensional membership vector set. Figure 7 The bar chart shows the multidimensional membership distribution of candidate cutting tools in this invention.

[0076] The business middleware calls the semantic rule inference engine to read the multidimensional membership vector set. A multidimensional performance acceptance threshold is set. Combining Figure 7 The figure shows a comparison of multidimensional membership degree distributions, reflecting the multidimensional performance evaluation results of candidate cutting tools. By examining the figure, which categorizes wear resistance, machining accuracy, and toughness as the horizontal axis, and using the modified final membership degree... The semantic rule reasoning engine performs cross-comparison and modern alignment determination based on the distribution state of the vertical axis.

[0077] Discovery such as Figure 7 The toughness of candidate tool A, represented by the medium-dark gray columnar structures, has a final membership degree of only [missing information]. ,like Figure 7 The final membership degree of the machining accuracy of candidate tool C, represented by the light gray columnar shape, is only [missing information]. ,like Figure 7 The candidate tool B, represented by the medium-gray bar, is penalized by exponential power operations, and its final membership degree is reduced to a certain value. . Reference Figure 7 As shown by the dotted lines, none of the three can simultaneously make the final membership degree of all dimensions greater than the multidimensional performance acceptance threshold. The semantic rule reasoning engine determines that a matching conflict has occurred, and writes the state conflict flag as true.

[0078] When the business middleware detects a true state conflict flag, it aborts the single-tool inference process and changes the state conflict flag to false. The middleware instantiates roughing and finishing task individuals based on the original part individual and issues different fuzzy modification instructions. For the roughing task individual, the machining accuracy operator strength is reset to the accuracy weakening operator. The toughness operator strength is maintained as that of the toughness strengthening operator. The semantic rule reasoning engine extracts candidate tool C through secondary deduction; for each finishing task, the toughness operator strength is reset to the toughness weakening operator. The strength of the wear-resistant operator remains the same as that of the wear-resistant strengthening operator. The semantic rule reasoning engine extracts candidate tool A through secondary deduction.

[0079] For the specific object query retrieval logic that extracts pointers to objects matching the specified characteristics from computer memory, those skilled in the art can implement it using existing graph database query languages. The graph structure traversal technique is a well-known technology in this field and will not be elaborated upon here. The output module extracts process coordination data including candidate tool C for roughing and candidate tool A for finishing, completes the structured selection result format encapsulation, and distributes it to external systems.

[0080] Based on the above embodiments and accompanying data, the following conclusions can be drawn regarding the implementation results of the technical solution of the present invention: according to Figure 6 (a) As can be seen from the data characterization of the toughness boundary translation of the tooth root region in this invention, the large module input triggers the positive compensation logic of the underlying reconstruction module. The right side fully subordinate boundary point is translated from 0.8 to 0.96, tightening the release standard that originally met 0.8 as having excellent toughness to 0.96. This filters out tool bodies that cannot withstand heavy impact at the physical level.

[0081] according to Figure 6 (b) Data characterization of the precision boundary contraction in the tooth surface region of this invention shows that the helix angle induces an axial deflection component. The algorithm reduces the transition tolerance range in the underlying mathematical model by shrinking the left absolute zero boundary point from 0.3 to 0.32588, forcing the system to examine the machining tolerance capability of the tool with more stringent precision convergence conditions. This indicates that the boundary dynamic reconstruction technology has broken through the technical bottleneck that static fuzzy sets cannot respond to the nonlinear time-varying load characteristics of mechanical dynamics.

[0082] according to Figure 7 Data characterization of the multidimensional membership distribution comparison diagram of the candidate cutting tools of the present invention shows that after the exponential operation of the degree fuzziness constraint operator is superimposed, the initial membership of the candidate cutting tool B, which was originally in a critical state, is quadratically attenuated to 0.49, directly exposing its physical shortcomings in comprehensive performance when facing extreme working conditions, causing it to fall below the baseline of the multidimensional performance acceptance threshold of 0.60.

[0083] at the same time, Figure 7 The results show that while candidate tool A far exceeds the threshold in terms of wear resistance and machining accuracy, its toughness is severely lacking; candidate tool C meets the toughness standard, but its machining accuracy is seriously insufficient. The bar chart distribution of the multidimensional membership vector set intuitively reflects that no single individual can overcome the multidimensional performance acceptance threshold, thus triggering the system's state conflict flag. Based on this, the business middleware dynamically derives execution instances, reducing the dimensionality of a single task and decoupling it into roughing and finishing sequences. This embodiment demonstrates that the present invention can effectively detect matching deadlocks caused by multivariate parameters and output process coordination data through secondary deduction of resetting operator strength, thus solving the inherently contradictory physical constraint problem between extreme impact resistance requirements and extreme surface quality requirements from an algorithmic mechanism perspective.

Claims

1. A method for intelligent tool selection based on fuzzy ontology, characterized in that, Includes the following steps: The system receives the macroscopic geometric parameters of the target gear and extracts the nominal parameters of the cutting tool. It then performs an instantiation operation in the fuzzy ontology knowledge base and defines the microscopic feature classification. The basic trapezoidal membership function model is invoked, and boundary compensation coefficients and compression ratio factors are generated by combining the macroscopic geometric parameters and the microscopic feature classification. The boundary compensation coefficients and compression ratio factors are used to perform boundary reconstruction calculation on the basic trapezoidal membership function model, and the reconstructed trapezoidal membership function is output. Based on the reconstructed trapezoidal membership function and degree fuzzy constraint operator adjustment model, the nominal parameters of the tool are processed, and a multidimensional membership vector set is output. Read the multidimensional membership vector set and determine the matching status. When a matching conflict occurs, generate a task individual, issue a fuzzy modification instruction to the task individual, and perform secondary logical deduction based on semantic rules. Extract and output the tool selection results after the secondary logic deduction to complete the intelligent tool selection for the target gear.

2. The intelligent tool selection method based on fuzzy ontology according to claim 1, characterized in that, The macroscopic geometric parameters include module, pressure angle and helix angle; the nominal parameters of the cutting tool include wear resistance, hot hardness, machining accuracy and toughness; and the microscopic feature classification includes tooth tip region, tooth surface region and tooth root region. The step of extracting the nominal parameters of the cutting tool specifically involves extracting the nominal parameters of the cutting standard data of the candidate cutting tool, writing the nominal parameters of the cutting tool as data attributes into the corresponding ontology instance, and storing them in the fuzzy ontology knowledge base.

3. The intelligent tool selection method based on fuzzy ontology according to claim 2, characterized in that, The specific steps for outputting the reconstructed trapezoidal membership function are as follows: The basic trapezoidal membership function model is invoked to extract the modulus and helix angle from the macroscopic geometric parameters, and the tooth root region from the microscopic feature classification is extracted. Based on the module and the helix angle, the boundary compensation coefficient and the compressibility factor are generated for the tooth root region; Using the boundary compensation coefficient and the compression factor, the initial boundary point coordinates of the basic trapezoidal membership function model are translated and contracted to calculate the reconstructed trapezoidal membership function.

4. The intelligent tool selection method based on fuzzy ontology according to claim 3, characterized in that, The specific steps for calculating the translation and contraction of the initial boundary point coordinates of the basic trapezoidal membership function model are as follows: The product of the first rigidity compensation coefficient and the modulus is calculated as the impact physics compensation increment. The impact physics compensation increment is added to the right fully membership boundary point and the right absolute zero boundary point of the basic trapezoidal membership function model to obtain the theoretical translation point of the right fully membership boundary without extreme value constraints and the theoretical translation point of the right absolute zero boundary without extreme value constraints. The minimum value between the theoretical translation point of the right-side fully subordinate boundary without extreme value constraints and the upper limit approximation safety value of the truncation constraint is taken as the reconstructed right-side fully subordinate boundary point. The minimum value between the theoretical translation point of the right-side absolute zero boundary without extreme value constraints and the upper limit extreme value of the normalized universe interval is taken as the reconstructed right-side absolute zero boundary point, thus completing the translation calculation of the initial boundary point coordinates. The product of the second rigidity compensation coefficient and the absolute value of the sine function calculation result of the helix angle is used as the vibration attenuation compensation amount. The product of the maximum compressibility factor and the initial length of the left transition interval of the basic trapezoidal membership function model is used as the maximum allowable interval shrinkage physical threshold. The minimum value between the vibration attenuation compensation amount and the maximum allowable interval shrinkage physical threshold is extracted as the left comprehensive scaling amount. The left absolute zero boundary point is obtained by adding the left comprehensive scaling amount to the left absolute zero boundary point of the basic trapezoidal membership function model, and the left complete membership boundary point is obtained by subtracting the left comprehensive scaling amount from the left complete membership boundary point of the basic trapezoidal membership function model, thus completing the shrinkage calculation of the initial boundary point coordinates.

5. The intelligent tool selection method based on fuzzy ontology according to claim 2, characterized in that, The specific steps for outputting the multidimensional membership vector set are as follows: Substitute the tool nominal parameters into the reconstructed trapezoidal membership function to perform algebraic calculations to obtain the initial membership degree value. Check whether the tool nominal parameters have valid values. If there are missing values ​​or non-numeric identifiers, force the corresponding initial membership degree value to be assigned a logical extreme value constant of zero. The degree fuzzy constraint operator adjustment model is invoked, and the modulus in the macroscopic geometric parameters is processed using a linear interpolation mapping mechanism to obtain a fixed real number as the operator intensity of the tooth root region; The absolute value of the helix angle in the macroscopic geometric parameters is processed using the linear interpolation mapping mechanism to obtain a fixed real number as the operator strength of the tooth surface region; Set the base to the initial membership value and the exponent to the operator strength, perform a power operation on the base to generate the final modified membership, and combine them to generate the multidimensional membership vector set.

6. The intelligent tool selection method based on fuzzy ontology according to claim 5, characterized in that, The specific steps for reading the multidimensional membership vector set, determining the matching status, and generating task individuals when a matching conflict occurs are as follows: Read the multidimensional membership vector set and extract the final membership degree under the corresponding evaluation dimension; The final membership degree in the multidimensional membership vector set is cross-compared with the set multidimensional performance acceptance threshold to determine whether the corresponding tool individual meets the multidimensional performance acceptance threshold condition. If the final membership degree of the corresponding tool individual in all dimensions is greater than the multidimensional performance acceptance threshold, then it is determined to be a successful match. If no individual tool in the multidimensional membership vector set can simultaneously satisfy the multidimensional performance acceptance threshold condition, then the matching conflict is determined to have occurred. When the matching conflict occurs, a state conflict identifier is written to the original part individual representing the target gear, the state conflict identifier is monitored to trigger the operation intervention, the state conflict identifier is erased, and a rough machining task individual and a fine machining task individual are instantiated based on the original part individual as the task individual.

7. The intelligent tool selection method based on fuzzy ontology according to claim 6, characterized in that, The specific steps of issuing fuzzy modification instructions to the task individual and performing secondary logical deduction based on semantic rules are as follows: For the individual roughing task, the fuzzy modification command that weakens precision and strengthens toughness is issued to reset the corresponding operator strength; For the individual finishing task, the fuzzy modification command that weakens toughness and strengthens wear resistance is issued to reset the corresponding operator strength; Using the reset operator strength as the new exponent, the base exponentiation operation is re-executed to generate the secondary modified membership degree, and the matching state determination logic is re-executed to extract the first tool subset and the second tool subset that meet the conditions in sequence. For candidates in the first and second tool subsets, the multidimensional weighted Euclidean distance evaluation logic is invoked: The difference between the preset ideal membership degree extreme value constant and the corresponding secondary modified membership degree is used as the physical performance difference. The product of the squares of the physical performance differences under all evaluation dimensions and the corresponding weight coefficients is summed, and the square root operation is performed on the summation result to obtain the comprehensive deviation distance. The individual with the smallest comprehensive deviation distance is extracted as the final physical output to match the corresponding tool entity.

8. The intelligent tool selection method based on fuzzy ontology according to claim 4, characterized in that, When calling the basic trapezoidal membership function model, it also includes setting the physical tolerance of the transition zone, regardless of whether it is for the initial boundary point coordinates or the reconstructed boundary point coordinates: When it is determined that the length of the left transition interval being processed is less than the physical tolerance of the transition interval, the corresponding left transition interval is downgraded to a step function. That is, when the input parameter is less than or equal to the critical threshold of the corresponding left interval, the function value output is zero, and when it is greater than the critical threshold, the function value output is one. When it is determined that the length of the right transition interval being processed is less than the physical tolerance of the transition interval, the corresponding right transition interval is downgraded to a step function. That is, when the input parameter is less than the critical threshold of the corresponding right interval, the function value output is one, and when it is greater than or equal to the critical threshold, the function value output is zero.

9. The intelligent tool selection method based on fuzzy ontology according to claim 2, characterized in that, The specific steps for writing the tool nominal parameters as data attributes into the corresponding ontology instance are as follows: Obtain the minimum and maximum values ​​of the tool nominal parameters set within the current candidate range, calculate the range between the maximum value and the minimum value, and set a physical discrimination threshold. When the range value is less than the physical discrimination threshold, the unbiased median is used as the normalized value. When the range value is not less than the physical discrimination threshold, the quotient of the difference between the extracted tool nominal parameter and the minimum value and the range value is calculated and used as the normalized value. The normalized value is written into the corresponding body instance as the nominal parameter of the tool.

10. The intelligent tool selection method based on fuzzy ontology according to claim 7, characterized in that, The specific steps for extracting and outputting the tool selection results after the secondary logical deduction are as follows: Receive rule parsing feedback data and extract process coordination data as the tool selection result; The process collaboration data includes roughing high-toughness tools and finishing high-wear-resistant tools corresponding to the task individual. The roughing high-toughness tool corresponds to the individual with the smallest overall deviation distance in the first tool subset, and the finishing high-wear-resistant tool corresponds to the individual with the smallest overall deviation distance in the second tool subset.

Images