A method for evaluating the trajectory interpolation performance of a numerical control system
By introducing the LSTM model and fuzzy comprehensive evaluation method, a multi-dimensional index system is constructed, which solves the comprehensive problem of trajectory interpolation performance evaluation of CNC system and realizes accurate evaluation and optimization of CNC system trajectory interpolation performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENYANG GOLDING NC & INTELLIGENCE TECH CO LTD
- Filing Date
- 2026-03-04
- Publication Date
- 2026-07-21
Smart Images
Figure CN122431259A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of CNC machining and intelligent manufacturing technology, specifically to a method for evaluating the trajectory interpolation performance of a CNC system, used to comprehensively evaluate the dynamic performance of a CNC system during trajectory interpolation. Background Technology
[0002] With the increasing demands for intelligent manufacturing and high-precision machining, CNC systems have become a crucial foundational equipment in modern manufacturing. Their performance directly impacts machine tool machining accuracy, production efficiency, and manufacturing costs. Trajectory interpolation, as a core component of CNC systems, is a key technology for achieving high-precision machining and smooth motion control, and it has a decisive influence on workpiece surface quality and overall machining accuracy.
[0003] In actual machining processes, CNC systems generally exhibit dynamic changes and nonlinear characteristics. Traditional interpolation performance evaluation methods that rely on a single index are insufficient to comprehensively reflect the system's true operating state and machining quality. In recent years, artificial intelligence technology has been increasingly applied to the field of CNC system performance analysis. Among these, Long Short-Term Memory (LSTM) networks have demonstrated good performance in processing time-series data and capturing dynamic features of the system. However, existing research largely focuses on predicting single performance indicators and lacks a trajectory interpolation performance evaluation scheme that combines time-series prediction models with multi-index comprehensive evaluation methods. Summary of the Invention
[0004] To address the shortcomings of existing CNC system trajectory interpolation performance evaluation methods in characterizing the dynamic characteristics of the trajectory interpolation process and their insufficient comprehensive evaluation capabilities, this invention proposes a CNC system trajectory interpolation function evaluation method based on a combination of Long Short-Term Memory (LSTM) networks and fuzzy comprehensive evaluation. This method enables a comprehensive evaluation of CNC system trajectory interpolation performance. This invention fully utilizes the temporal characteristics of the trajectory interpolation process and achieves a multi-performance index fusion evaluation method for CNC system trajectory interpolation function, thereby improving the accuracy and reliability of the evaluation results and providing effective support for CNC system performance optimization and machining quality improvement.
[0005] The technical solution adopted by the present invention to achieve the above objectives is as follows:
[0006] A method for evaluating the trajectory interpolation performance of a CNC system includes the following steps:
[0007] 1) Collect instruction trajectory data and actual feedback trajectory data of the CNC system during the machining process according to a fixed sampling period, construct a multi-dimensional time series trajectory dataset, and preprocess it;
[0008] 2) Train a long short-term memory network model using a sequence trajectory dataset, and use this model to predict trajectory tracking errors;
[0009] 3) Select the trajectory interpolation error sensitive area by observing the changes in trajectory curvature, and construct an evaluation index system within this area;
[0010] 4) Determine the weights of evaluation indicators by combining subjective and objective methods;
[0011] 5) The evaluation indicators are fuzzyened and combined with their corresponding weights to comprehensively evaluate the trajectory interpolation performance of the CNC system.
[0012] Step 2) specifically refers to:
[0013] Based on the deviation between the command trajectory and the actual feedback trajectory, trajectory tracking error data is constructed. The sequence trajectory data is used as the input to the long short-term memory network model, and the trajectory deviation is used as the output to train the model. The trained model is then used to predict new trajectory time series data to obtain the trajectory tracking error prediction results at the corresponding time.
[0014] The selection of the trajectory interpolation error sensitive area based on the change in trajectory curvature specifically involves:
[0015] Calculate the trajectory curvature using the first and second derivatives of the trajectory space vector. :
[0016] ;
[0017] Where r is a spatial trajectory vector composed of multi-axis trajectory data or multi-axis trajectory tracking error;
[0018] When the trajectory curvature satisfies At that time, the corresponding trajectory segment is determined as the trajectory interpolation error sensitive area, where This is the preset curvature threshold.
[0019] The evaluation index system includes:
[0020] Mean Indicator :
[0021] ;
[0022] in, The number of sampling points. Let be the trajectory tracking error value at the i-th sampling time. The sampling point number;
[0023] variance index :
[0024] ;
[0025] Speed Index :
[0026] ;
[0027] in, for t Time-tracking error function;
[0028] Acceleration index :
[0029] .
[0030] Step 4) includes the following steps:
[0031] 4.1) The Analytic Hierarchy Process (AHP) is used to subjectively assign weights to the trajectory interpolation performance evaluation indicators, resulting in a subjective weight vector. ;
[0032] 4.2) The entropy weight method is used to objectively assign weights to the trajectory interpolation performance evaluation index, resulting in an objective weight vector. ;
[0033] 4.3) The subjective and objective weights are weighted and fused to obtain the comprehensive weight vector of each evaluation indicator. :
[0034] ;
[0035] Where α is the weighted fusion coefficient.
[0036] Step 4.1) specifically refers to:
[0037] Construct a judgment matrix based on the importance relationships among the evaluation indicators. ,in, This represents the importance of the i-th evaluation indicator relative to the j-th evaluation indicator, determined by solving for the maximum eigenvalue λ of the judgment matrix. max The subjective weight vectors of each evaluation index are obtained by normalizing the eigenvectors and their corresponding feature vectors.
[0038] ;
[0039] The consistency of the obtained subjective weight vector is checked by using the consistency index CI and the consistency ratio CR. When CR meets the threshold, the subjective weight vector is judged to meet the consistency requirements.
[0040] Step 4.2) specifically refers to:
[0041] Suppose there are m evaluation samples, construct the original data matrix of evaluation indicators. For the matrix After normalization, the index ratio matrix is obtained. :
[0042] ;
[0043] Based on the information entropy theory, calculate the information entropy value of the j-th evaluation index. :
[0044] ;
[0045] in, This is a normalization constant;
[0046] Calculate the information utility value of the indicators to obtain the objective weights of each evaluation indicator:
[0047] .
[0048] Step 5) includes the following steps:
[0049] 5.1) Based on the requirements for evaluating trajectory interpolation performance, set an evaluation level set. ;
[0050] 5.2) The actual values of the evaluation indicators are mapped to the membership degrees of the corresponding evaluation levels, and a trapezoidal membership function is used to fuzzify each evaluation indicator, thereby constructing a fuzzy relation matrix for trajectory interpolation performance evaluation. ;
[0051] 5.3) Combine the weight vector With fuzzy relation matrix Perform fuzzy comprehensive calculations to obtain a comprehensive evaluation vector of trajectory interpolation performance. :
[0052] ;
[0053] 5.4) Based on comprehensive evaluation vector The evaluation level corresponding to the trajectory interpolation performance of the CNC system is determined based on the principle of maximum membership.
[0054] The trapezoidal membership function is:
[0055] ;
[0056] in, , , , This is the segmented threshold parameter.
[0057] The fuzzy relation matrix for evaluating trajectory interpolation performance for:
[0058] ;
[0059] Where, r ij This represents the membership degree of the i-th evaluation index to the j-th evaluation level.
[0060] The present invention has the following beneficial effects and advantages:
[0061] 1. This invention introduces a long short-term memory network model to predict trajectory tracking errors in a time series. This model can fully explore the temporal correlation and nonlinear dynamic characteristics of error changes during trajectory interpolation, and improves the ability to characterize the dynamic performance of the system compared with traditional static statistical analysis methods.
[0062] 2. This invention extracts the trajectory interpolation error sensitive area by using a trajectory curvature threshold, and focuses on analyzing key trajectory segments with large curvature changes. This avoids the information masking problem caused by averaging the entire trajectory, thereby enhancing the sensitivity of the evaluation results to key error segments.
[0063] 3. This invention constructs a multi-dimensional evaluation system that includes indicators such as mean error, variance of error, rate of change of error, and acceleration of error, which not only reflects the trajectory interpolation accuracy but also reflects the dynamic response characteristics of the system.
[0064] 4. This invention uses a combination of the analytic hierarchy process (AHP) and the entropy weight method to determine the weights of the evaluation indicators. While fully considering expert experience, it introduces a data-driven objective weight calculation method to reduce the bias caused by a single subjective or objective weighting and improve the rationality of the evaluation results. Attached Figure Description
[0065] Figure 1 Overall structural diagram for evaluating the trajectory interpolation performance of a CNC system;
[0066] Figure 2 LSTM tracking error prediction model structure diagram;
[0067] Figure 3 Structure diagram of the comprehensive evaluation model for trajectory interpolation performance of CNC system;
[0068] Figure 4 Schematic diagram of sensitive area extraction. Detailed Implementation
[0069] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.
[0070] like Figure 1 As shown, a method for evaluating the trajectory interpolation performance of a CNC system includes the following steps:
[0071] 1) Data acquisition: Acquire command trajectory data and actual feedback trajectory data of the CNC system during the machining process, including command position, actual position, speed and acceleration of each axis, and form a time series dataset according to a fixed sampling period;
[0072] 2) LSTM trajectory tracking error prediction: After preprocessing the trajectory data, a long short-term memory network model is constructed to perform time-series prediction of the trajectory tracking error during the trajectory interpolation process, and the prediction result of the trajectory tracking error is obtained.
[0073] 3) Construction of evaluation index system: Based on the changes in trajectory curvature, the trajectory regions with large curvature changes are extracted as the sensitive regions for trajectory interpolation error. The mean, maximum, and variance of trajectory tracking error, as well as the corresponding velocity and acceleration indicators, are selected to construct a trajectory interpolation performance evaluation index system.
[0074] 4) Determining the weight of indicators: The subjective weight of each evaluation indicator is determined by the analytic hierarchy process, while the objective weight of each evaluation indicator is determined by the entropy weight method. The comprehensive weight is obtained by weighted fusion to take into account both expert experience and data objectivity.
[0075] 5) Fuzzy comprehensive evaluation method: Based on the preset evaluation level set and membership function, each evaluation index is fuzzified to construct a fuzzy relation matrix, and fuzzy operation is performed in combination with the comprehensive weight to obtain the comprehensive evaluation result of the trajectory interpolation performance of the CNC system.
[0076] 1. Trajectory data acquisition and preprocessing
[0077] The CNC system's data acquisition interface collects command trajectory data and actual feedback trajectory data for each axis during machining. The trajectory data includes the commanded position, actual position, velocity, and acceleration information for each axis. During acquisition, the trajectory data is continuously sampled according to a preset fixed sampling period, forming a multi-dimensional time-series trajectory dataset. To improve data quality and the stability of subsequent processing, the collected raw trajectory data undergoes preprocessing, including outlier removal, missing data correction, and normalization. A time-series trajectory sample is then constructed in chronological order to provide input data for subsequent trajectory tracking error prediction.
[0078] 2. LSTM-based trajectory tracking error prediction
[0079] like Figure 2 As shown, after completing trajectory data acquisition and preprocessing, trajectory tracking error data is constructed based on the deviation between the command trajectory and the actual feedback trajectory. Using the preprocessed multidimensional trajectory time series data as input, a trajectory tracking error prediction model based on a long short-term memory neural network is established.
[0080] By training the LSTM model on historical trajectory data, the model learns the temporal correlation and dynamic changes of trajectory tracking errors during trajectory interpolation. After training, new trajectory time series data is input into the LSTM model to obtain the trajectory tracking error prediction results at the corresponding time points, thereby enabling the prediction of the trend of trajectory interpolation error changes.
[0081] 3. Construction of a performance evaluation index system for trajectory interpolation
[0082] Within the extracted trajectory interpolation error sensitive area, statistical analysis is performed on the trajectory tracking error prediction results. Combined with motion state information during the trajectory interpolation process, a trajectory interpolation performance evaluation index system is constructed. This system includes error-related indicators reflecting trajectory interpolation accuracy, as well as velocity and acceleration indicators reflecting the system's dynamic characteristics. This achieves a multi-dimensional comprehensive characterization of the CNC system's trajectory interpolation performance, providing a foundation for subsequent evaluation models.
[0083] like Figure 4 As shown, the trajectory curvature is calculated based on the first and second derivatives of the trajectory space vector, and its curvature satisfies:
[0084]
[0085] Where r is a spatial trajectory vector composed of multi-axis trajectory data or multi-axis trajectory tracking errors.
[0086] When the trajectory curvature satisfies:
[0087]
[0088] At that time, the corresponding trajectory segment is determined as the trajectory interpolation error sensitive area, where K th This is the preset curvature threshold.
[0089] The statistical indicators of trajectory tracking error include the mean and variance, which are calculated according to the following conditions:
[0090]
[0091]
[0092] The velocity and acceleration indices are calculated using the first and second differences of the trajectory tracking error, respectively, and satisfy the following conditions:
[0093]
[0094] 4. Methods for determining the weights of evaluation indicators
[0095] To ensure the rationality and objectivity of the trajectory interpolation performance evaluation results, the weights of the evaluation indicators are determined using a combination of subjective and objective methods.
[0096] First, the Analytic Hierarchy Process (AHP) is used to subjectively assign weights to the trajectory interpolation performance evaluation indicators. A judgment matrix is then constructed based on the importance relationships among the evaluation indicators. , where a ij This represents the importance of the i-th evaluation indicator relative to the j-th evaluation indicator. The maximum eigenvalue λ of the judgment matrix is calculated. max The subjective weight vectors of each evaluation index are obtained by normalizing the corresponding feature vectors.
[0097]
[0098] To ensure the rationality of the judgment matrix, a consistency check is performed on it. The consistency index CI and the consistency ratio CR are expressed as follows:
[0099]
[0100] When CR < 0.1, the judgment matrix is considered to meet the consistency requirement.
[0101] Meanwhile, to reduce the influence of subjective factors on the evaluation results, the entropy weight method is introduced to objectively assign weights to the evaluation indicators. Assuming there are m sets of evaluation samples, an original data matrix of evaluation indicators is constructed. After normalizing the matrix, the index ratio matrix is obtained:
[0102]
[0103] According to the information entropy theory, calculate the information entropy value of the j-th evaluation index:
[0104]
[0105] Further calculate the information utility value of the indicators and obtain the objective weights of each evaluation indicator:
[0106]
[0107] Based on this, the subjective and objective weights are weighted and fused to obtain the comprehensive weight vector of each evaluation indicator:
[0108]
[0109] Wherein, α is the weight fusion coefficient, used to adjust the proportion of subjective weight and objective weight in the overall weight.
[0110] 5. Fuzzy Comprehensive Evaluation Model
[0111] like Figure 3 As shown, after obtaining the comprehensive weights of each evaluation index, a fuzzy comprehensive evaluation model for the trajectory interpolation performance of the CNC system is constructed.
[0112] First, based on the requirements for evaluating trajectory interpolation performance, the evaluation level set is set as follows:
[0113]
[0114] The evaluation level set is used to characterize different evaluation levels of the trajectory interpolation performance of the CNC system, from excellent to poor.
[0115] Secondly, to map the actual values of the evaluation indicators to the membership degrees of the corresponding evaluation levels, a trapezoidal membership function is used to fuzzify each evaluation indicator. Let the value of the evaluation indicator be x, and its membership function be expressed as:
[0116]
[0117] This allows us to construct a fuzzy relation matrix for evaluating trajectory interpolation performance:
[0118]
[0119] Where, r ij This represents the membership degree of the i-th evaluation index to the j-th evaluation level.
[0120] Finally, the comprehensive weight vector W and the fuzzy relation matrix R are subjected to fuzzy comprehensive operation to obtain the comprehensive evaluation vector of trajectory interpolation performance:
[0121]
[0122] The evaluation level corresponding to the trajectory interpolation performance of the CNC system is determined based on the principle of maximum membership.
Claims
1. A method for evaluating the trajectory interpolation performance of a CNC system, characterized in that, Includes the following steps: 1) Collect instruction trajectory data and actual feedback trajectory data of the CNC system during the machining process according to a fixed sampling period, construct a multi-dimensional time series trajectory dataset, and preprocess it; 2) Train a long short-term memory network model using a sequence trajectory dataset, and use this model to predict trajectory tracking errors; 3) Select the trajectory interpolation error sensitive area by observing the changes in trajectory curvature, and construct an evaluation index system within this area; 4) Determine the weights of evaluation indicators by combining subjective and objective methods; 5) The evaluation indicators are fuzzyened and combined with their corresponding weights to comprehensively evaluate the trajectory interpolation performance of the CNC system.
2. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 1, characterized in that, Step 2) specifically refers to: Based on the deviation between the command trajectory and the actual feedback trajectory, trajectory tracking error data is constructed. The sequence trajectory data is used as the input to the long short-term memory network model, and the trajectory deviation is used as the output to train the model. The trained model is then used to predict new trajectory time series data to obtain the trajectory tracking error prediction results at the corresponding time.
3. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 1, characterized in that, The selection of the trajectory interpolation error sensitive area based on the change in trajectory curvature specifically involves: Calculate the trajectory curvature using the first and second derivatives of the trajectory space vector. : ； Where r is a spatial trajectory vector composed of multi-axis trajectory data or multi-axis trajectory tracking error; When the trajectory curvature satisfies At that time, the corresponding trajectory segment is determined as the trajectory interpolation error sensitive area, where This is the preset curvature threshold.
4. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 1, characterized in that, The evaluation index system includes: Mean Indicator : ； in, The number of sampling points. Let be the trajectory tracking error value at the i-th sampling time. The sampling point number; variance index : ； Speed Index : ； in, for t Time-tracking error function; Acceleration index : 。 5. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 1, characterized in that, Step 4) includes the following steps: 4.1) The Analytic Hierarchy Process (AHP) is used to subjectively assign weights to the trajectory interpolation performance evaluation indicators, resulting in a subjective weight vector. ; 4.2) The entropy weight method is used to objectively assign weights to the trajectory interpolation performance evaluation index, resulting in an objective weight vector. ; 4.3) The subjective and objective weights are weighted and fused to obtain the comprehensive weight vector of each evaluation indicator. : ； Where α is the weighted fusion coefficient.
6. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 5, characterized in that, Step 4.1) specifically refers to: Construct a judgment matrix based on the importance relationships among the evaluation indicators. ,in, This represents the importance of the i-th evaluation indicator relative to the j-th evaluation indicator, determined by solving for the maximum eigenvalue λ of the judgment matrix. max The subjective weight vectors of each evaluation index are obtained by normalizing the eigenvectors and their corresponding feature vectors. ； The consistency of the obtained subjective weight vector is checked by using the consistency index CI and the consistency ratio CR. When CR meets the threshold, the subjective weight vector is judged to meet the consistency requirements.
7. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 5, characterized in that, Step 4.2) specifically refers to: Suppose there are m evaluation samples, construct the original data matrix of evaluation indicators. For the matrix After normalization, the index ratio matrix is obtained. : ； Based on the information entropy theory, calculate the information entropy value of the j-th evaluation index. : ； in, This is a normalization constant; Calculate the information utility value of the indicators to obtain the objective weights of each evaluation indicator: 。 8. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 1, characterized in that, Step 5) Includes the following steps: 5.1) Based on the requirements for evaluating trajectory interpolation performance, set an evaluation level set. ; 5.2) The actual values of the evaluation indicators are mapped to the membership degrees of the corresponding evaluation levels, and a trapezoidal membership function is used to fuzzify each evaluation indicator, thereby constructing a fuzzy relation matrix for trajectory interpolation performance evaluation. ; 5.3) Combine the weight vector With fuzzy relation matrix Perform fuzzy comprehensive calculations to obtain a comprehensive evaluation vector of trajectory interpolation performance. : ； 5.4) Based on comprehensive evaluation vector The evaluation level corresponding to the trajectory interpolation performance of the CNC system is determined based on the principle of maximum membership.
9. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 8, characterized in that, The trapezoidal membership function is: ； in, , , , This is the segmented threshold parameter.
10. The method for evaluating the trajectory interpolation performance of a CNC system according to claim 8, characterized in that, The fuzzy relation matrix for evaluating trajectory interpolation performance for: ； Where, r ij This represents the membership degree of the i-th evaluation index to the j-th evaluation level.