A diesel engine connecting rod assembly deformation prediction method, device, equipment and medium

Abstract

This invention relates to a method, apparatus, equipment, and medium for predicting deformation during the assembly of diesel engine connecting rods, belonging to the field of mechanical assembly technology. The method acquires multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods. Based on assembly process knowledge, the multi-dimensional assembly data is weighted to obtain a weighted input vector set. A pre-trained Bayesian network model is used to infer the deformation prediction probability distribution from the weighted input vector set. Statistical analysis is performed on the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics. Key deformation risk factors are identified based on these characteristics. A diesel engine connecting rod assembly deformation prediction report is generated based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors. This invention improves the assembly quality and product reliability of marine diesel engine connecting rods.

Description

Technical Field

[0001] This invention relates to the field of mechanical assembly technology, and in particular to a method, device, equipment and medium for predicting deformation during diesel engine connecting rod assembly. Background Technology

[0002] Traditional prediction of diesel engine connecting rod assembly deformation mainly relies on manual experience or simple statistical models (such as linear regression), and makes rough assessments using limited manual measurement data or historical records.

[0003] However, traditional methods for predicting diesel engine connecting rod assembly deformation lack comprehensive analysis and probabilistic reasoning of multi-dimensional assembly data. As a result, they suffer from problems such as low prediction accuracy, inability to identify early risks, insufficient support for assembly process optimization, and difficulty in handling uncertainties. Consequently, diesel engine connecting rod assembly quality fluctuates greatly, and product reliability is limited. Summary of the Invention

[0004] In view of this, it is necessary to provide a method, device, equipment and medium for predicting deformation of diesel engine connecting rod assembly, so as to solve the technical problems of low prediction accuracy of diesel engine connecting rod assembly, inability to identify risks in the early stage and insufficient support for process optimization.

[0005] To address the aforementioned problems, in a first aspect, the present invention provides a method for predicting deformation during diesel engine connecting rod assembly, comprising: Multi-dimensional assembly data during the assembly process of marine diesel engine connecting rod is obtained, and the multi-dimensional assembly data is weighted based on assembly process knowledge to obtain a weighted input vector set. The weighted input vector set is inferred based on a pre-trained Bayesian network model to obtain the deformation prediction probability distribution. Statistical analysis is performed on the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics, and key deformation risk factors are identified based on the deformation trend characteristics and distribution dispersion characteristics. Based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors, a deformation prediction report for diesel engine connecting rod assembly is generated.

[0006] In one possible implementation, the weighting of the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set includes: Obtain the deformation influencing factor and the quantitative value of the degree of influence of the deformation influencing factor in the connecting rod assembly process of the assembly process knowledge; Calculate the feature matching degree between the multi-dimensional assembly data and the deformation influencing factors; Calculate the weight value of each dimension of the assembly data based on the feature matching degree and the quantification value of the degree of influence. The assembly data for each dimension is weighted based on the weight values ​​to generate a weighted input vector set.

[0007] In one possible implementation, the Bayesian network model includes a component preprocessing layer, an assembly execution layer, and a deformation generation layer. The pre-trained Bayesian network model infers from the weighted input vector set to obtain a deformation prediction probability distribution, including: Based on the component preprocessing layer, feature extraction is performed on the weighted input vector set to obtain component features; Based on the assembly execution layer, the component features are assembled and mapped to obtain an assembly feature vector; Based on the deformation generation layer, deformation probability inference is performed on the component features and assembly feature vectors, and the deformation prediction probability distribution is output.

[0008] In one possible implementation, the deformation prediction probability distribution is: , in, For the deformation prediction probability distribution, For assembly deformation state variables, For component feature vectors, For assembly feature vectors, For the overall parameters of the Bayesian network model, The number of nodes in the deformation generation layer. For component feature weights, For assembly feature weights, For the corresponding number The component feature vectors of each node, For the corresponding number Assembly feature vectors of each node For the first The average value of the deformation state conditions of each node. For the first The conditional variance of the deformation state of each node. Let be the prior distribution parameters of the mean value of the deformation state conditions. Let be the prior distribution parameter of the variance of the deformation state condition. Let be the probability density function of a normal distribution. Let be the probability density function of the gamma distribution. For joint integral operators, Let be the probability density function of the conditional variance of the deformable state. This represents the conditional probability distribution of the assembly deformation state variables.

[0009] In one possible implementation, the statistical analysis of the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics includes: A preset sliding window is used to sample the deformation prediction probability distribution to obtain a probability distribution subsequence; The mean of the probability distribution subsequences is calculated to construct a deformed probability time series sequence; A fitting function is obtained by performing trend fitting on the deformation probability time series, and the feature parameters of the fitting function are used as deformation trend features. Calculate the residual between the deformation probability time series and the fitting function at each sampling point, calculate the confidence interval width of the deformation prediction probability distribution based on the residual, calculate the information entropy of the deformation prediction probability distribution, and determine the distribution dispersion characteristics based on the confidence interval width and information entropy.

[0010] In one possible implementation, the identification of key deformation risk factors based on the deformation trend characteristics and distribution dispersion characteristics includes: Calculate the correlation between the multi-dimensional assembly data and the deformation trend characteristics to obtain a first correlation value; Calculate the correlation degree between the multi-dimensional assembly data and the distribution dispersion feature to obtain a second correlation degree value; The multi-dimensional assembly data is filtered based on a preset first correlation threshold and a first correlation value to obtain a first type of candidate data; The multi-dimensional assembly data is filtered based on a preset second correlation threshold and a second correlation value to obtain a second type of candidate data; The first type of candidate data and the second type of candidate data are integrated to obtain key assembly data, and the assembly process parameters corresponding to the key assembly data are obtained, and the assembly process parameters are used as key deformation risk factors.

[0011] In one possible implementation, the diesel engine connecting rod assembly deformation prediction report includes deformation prediction results, risk quantification indicators, and process optimization suggestions.

[0012] Secondly, the present invention also provides a diesel engine connecting rod assembly deformation prediction device, comprising: The data acquisition module is used to acquire multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods, and to weight the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set. The inference module is used to infer the weighted input vector set based on a pre-trained Bayesian network model to obtain the deformation prediction probability distribution. The risk factor identification module is used to perform statistical analysis on the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics, and to identify key deformation risk factors based on the deformation trend characteristics and distribution dispersion characteristics. The prediction report generation module is used to generate a diesel engine connecting rod assembly deformation prediction report based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors.

[0013] Thirdly, the present invention also provides a diesel engine connecting rod assembly deformation prediction device, comprising: a processor and a memory; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the diesel engine connecting rod assembly deformation prediction method as described above.

[0014] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in the diesel engine connecting rod assembly deformation prediction method described in any one of the above-mentioned method items.

[0015] The beneficial effects of this invention are as follows: It acquires multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods; weights the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set; assigns prior weights to the multi-dimensional assembly data through assembly process knowledge, thus optimizing the quality and relevance of the data input; it infers from the weighted input vector set based on a pre-trained Bayesian network model to obtain a deformation prediction probability distribution; through probabilistic inference of the Bayesian network model, it obtains the deformation prediction probability distribution, effectively handling the uncertainty in the assembly process and improving prediction accuracy; and it performs statistical analysis on the deformation prediction probability distribution. This method obtains deformation trend characteristics and distribution dispersion characteristics, identifies key deformation risk factors based on these characteristics, and achieves early risk identification and reliability assessment through statistical analysis of the deformation prediction probability distribution. By identifying key deformation risk factors, it provides a basis for targeted optimization of the assembly process. Based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors, a deformation prediction report for diesel engine connecting rod assembly is generated. This overcomes the shortcomings of traditional methods, such as low prediction accuracy, inability to identify early risks, and insufficient support for process optimization, thereby improving the assembly quality and product reliability of marine diesel engine connecting rods. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart of an embodiment of the diesel engine connecting rod assembly deformation prediction method provided by the present invention; Figure 2 A flowchart illustrating the statistical analysis of the diesel engine connecting rod assembly deformation prediction method provided by this invention; Figure 3 A flowchart illustrating the identification of key deformation risk factors in the diesel engine connecting rod assembly deformation prediction method provided by the present invention. Figure 4 A schematic diagram of an embodiment of the diesel engine connecting rod assembly deformation prediction device provided by the present invention; Figure 5 This is a schematic diagram of an embodiment of the diesel engine connecting rod assembly deformation prediction device provided by the present invention. Detailed Implementation

[0018] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0019] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0020] This invention discloses a method, apparatus, device, and medium for predicting deformation of diesel engine connecting rod assembly, which can be used in a computer. The method, apparatus, or computer-readable storage medium involved in this invention can be integrated with the aforementioned equipment or be relatively independent.

[0021] One specific embodiment of the present invention discloses a method for predicting deformation during diesel engine connecting rod assembly, which can be executed by a computer, specifically by one or more processors of the computer. For example... Figure 1 As shown, the method for predicting deformation of diesel engine connecting rod assembly includes: S101. Obtain multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods, and weight the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set. It should be noted that by collecting multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods, and combining this data with prior weights based on assembly process knowledge to form a weighted input vector set, the quality and relevance of the data input were optimized.

[0022] S102. Based on the pre-trained Bayesian network model, reason about the weighted input vector set to obtain the deformation prediction probability distribution; It should be noted that the Bayesian network model includes a component preprocessing layer, an assembly execution layer, and a deformation generation layer. By performing probabilistic inference through the Bayesian network model, the deformation prediction probability distribution is obtained, which effectively handles the uncertainty in the assembly process and improves the prediction accuracy.

[0023] S103. Perform statistical analysis on the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics, and identify key deformation risk factors based on deformation trend characteristics and distribution dispersion characteristics. It should be noted that, through statistical analysis of the deformation prediction probability distribution, early risk identification and reliability assessment were achieved. By identifying key deformation risk factors, a basis was provided for targeted optimization of the assembly process.

[0024] S104. Generate a diesel engine connecting rod assembly deformation prediction report based on deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors.

[0025] In some embodiments, in step S101, multi-dimensional assembly data is acquired during the assembly process of marine diesel engine connecting rods. By deploying multiple sensors (such as optical scanners, displacement sensors, or force sensors) at key workstations throughout the entire process of marine diesel engine connecting rod assembly, geometric parameters of connecting rod components, positioning parameters of assembly tooling, and assembly operation parameters are collected in real time to form raw assembly data. The raw assembly data is then classified and integrated according to the assembly process sequence to form multi-dimensional assembly data. This multi-dimensional assembly data covers multiple aspects such as space, time, and process parameters, achieving data comprehensiveness and representativeness, and providing a reliable guarantee for deformation prediction. Key workstations may include six core workstations: parts receiving inspection station, large-end hole precision machining station, large-end cap assembly station, bolt tightening station, small-end bushing press-fitting station, and overall geometric accuracy inspection station. Each workstation is equipped with corresponding sensors according to its function: the parts receiving inspection station collects data such as the connecting rod diameter (design value). Collection range ), inner diameter of the large end hole (design value) Collection range ), coaxiality of the large end hole (requirement) Small end hole roundness (requirements) Four geometric parameters of the connecting rod components were collected, including those for connecting rods and bushings. The data was collected once per batch of 20 connecting rods, generating 200 data points each time to fit the actual parameter values. Data on the tooling positioning pins was collected at the large end cap assembly station and the small end bushing pressing station. Deviation (allowable range) ), Deviation (allowable range) ), Deviation (allowable range) ) and bushing press-fit positioning reference deviation (allowable range) Four assembly tooling positioning parameters, including ( ), are collected once per assembly action (i.e., once per connecting rod assembly), with the tooling number recorded synchronously; the bolt tightening torque (design value) is collected at the bolt tightening station. Collection range ), Fastening angle (design value) Collection range Torque rise rate (requirement) Three assembly operation parameters, including [parameter 1], were collected at a frequency of 10Hz (i.e., 10 sets of data per second) until the bolt tightening process was completed (each process lasted approximately 15 seconds, with a total of 130 sets of data collected); the overall geometric accuracy inspection station collected the center distance between the large and small end holes of the connecting rod assembly (design value). Collection range ), shaft straightness (requirements) Two comprehensive geometric parameters, namely, the sampling frequency is 1 out of every 10 connecting rods, with 3 cross-sectional data points collected for each rod to calculate the average value. All sensors can be connected to the data acquisition terminal via industrial Ethernet (Profinet protocol). The collected parameter values, acquisition timestamps, sensor numbers, workstation numbers, connecting rod batch numbers, and other information are stored as raw assembly data. The raw assembly data is classified and integrated according to the standard assembly process sequence of marine diesel engine connecting rods (e.g., "parts warehousing inspection, large end hole precision machining, large end cap assembly, bolt tightening, small end bushing press-fitting, overall geometric accuracy inspection"). The timestamps of the raw data are associated with the execution time intervals of each process through the process time of the data acquisition terminal. Raw data whose timestamps fall within the corresponding intervals are classified into the data group for that process. Parameter classification is performed for each process data group. For example, the "large end cap assembly process data group" includes all positioning pins collected by laser displacement sensors under that process. The deviation data, specifically the "bolt tightening process data group," includes torque, rotation angle, and torque rise rate data. Preprocessing similar parameters allows for the construction of a multi-dimensional assembly data matrix based on "process sequence - parameter type." The matrix row index represents a unified time sampling point (rounded to the nearest 1 second, covering the entire assembly cycle, approximately 300 sampling points), and the column index represents the parameter items for each process (19 parameters, i.e., 19 dimensions). Each matrix element represents the preprocessed value of the corresponding sampling point and parameter, forming a dimensional matrix. Multi-dimensional assembly data.

[0026] Based on assembly process knowledge, multi-dimensional assembly data is weighted to obtain a weighted input vector set. The deformation influencing factors and their quantified influence values ​​for the connecting rod assembly process are obtained from the assembly process knowledge. The feature matching degree between the multi-dimensional assembly data and the deformation influencing factors is calculated. Based on the feature matching degree and the quantified influence value, the weight value of each dimension of assembly data is calculated. The assembly data for each dimension is then weighted based on these weight values ​​to generate a weighted input vector set. Deformation influencing factors (such as geometric deviations of connecting rod components, positioning errors of assembly tooling, and pressure parameters during assembly operations) corresponding to each connecting rod assembly process are retrieved from a pre-set assembly process knowledge base. Using the analytic hierarchy process (AHP) combined with expert scoring results, the influence of each deformation influencing factor on assembly deformation is quantified into a specific value between 0 and 1 (the stronger the influence, the closer the quantified value is to 1). The cosine similarity algorithm is used to calculate the features of each dimension of assembly data. The similarity between the vector and the feature vectors of each deformation influencing factor is used to obtain the feature matching degree between each dimension's assembled data and the corresponding deformation influencing factor (the matching degree ranges from 0 to 1, with the closer the value is to 1, the stronger the association between the dimension's data and the influencing factor). The weight value of each dimension's assembled data is calculated by weighted product summation. The feature matching degree of each dimension's assembled data and each deformation influencing factor is multiplied by the quantified influence value of the corresponding deformation influencing factor, and then all product results are summed. The summation results of each dimension's data are normalized (so that the sum of the weight values ​​of all dimension's data is 1) to obtain the weight value of each dimension's assembled data. According to the preset arrangement order of the multi-dimensional assembled data, the original value of each dimension's assembled data is multiplied by the corresponding weight value, and all weighted values ​​are arranged in the original dimension order to form a weighted input vector set, which provides an optimization basis for subsequent Bayesian network inference.

[0027] In some implementations, in step S102, the weighted input vector set is inferred based on a pre-trained Bayesian network model to obtain the deformation prediction probability distribution. The Bayesian network model includes a component preprocessing layer, an assembly execution layer, and a deformation generation layer. The pre-trained Bayesian network model adopts a three-level node network structure consisting of the component preprocessing layer, assembly execution layer, and deformation generation layer. Based on the component preprocessing layer, features are extracted from the weighted input vector set to obtain component features; based on the assembly execution layer, assembly mapping is performed on the component features to obtain assembly feature vectors; based on the deformation generation layer, deformation probability inference is performed on the component features and assembly feature vectors to output the deformation prediction probability distribution. The three-level node network structure adopted by the pre-trained Bayesian network model includes a component preprocessing layer containing 12 nodes, each corresponding to a geometric component of the link. Parameters (such as connecting rod diameter deviation, large end hole coaxiality, small end hole roundness), material parameters (such as hardness, elastic modulus), and preprocessing state parameters (such as cleanliness, surface roughness) are used. Each node can perform feature transformation on the corresponding dimension of the weighted input vector set through the Sigmoid activation function to extract key features of the parts. The assembly execution layer contains 8 nodes. Based on the assembly process sequence of marine diesel engine connecting rods (such as large end cap assembly, bolt tightening, small end bushing pressing, and overall inspection) and tooling positioning constraints (such as locating pin position and clamping force threshold), an attention mechanism is used to associate and map the parts features with the tooling positioning parameters and operating force parameters in the assembly process. By calculating the mutual information value between the parts features and the parameters of each assembly process (with a threshold set to 0.6), key related items are filtered. The fully connected layer integrates them into a dimension of The model's assembly feature vector is generated; the deformation generation layer contains 6 nodes, corresponding to 6 typical states of assembly deformation (no deformation, slight deformation (0-0.1mm), mild deformation (0.1-0.2mm), moderate deformation (0.2-0.3mm), severe deformation (0.3-0.4mm), and serious deformation (>0.4mm)). When constructing the training dataset for the model, the real assembly data comes from 1400 sets of full-process data of connecting rod assembly collected by multiple sensors (covering different batches of connecting rod parts, 3 mainstream tooling models, and 5 typical operating conditions). The simulation data is obtained through ANSYS Mechanica. The software generates a dynamic model of the connecting rod assembly. The simulation scenarios include eight common interference conditions such as tooling positioning error fluctuations and assembly force deviations, totaling 600 sets. Real and simulation data are randomly mixed at a 7:3 ratio (i.e., every 14 sets of real data are paired with 6 sets of simulation data, resulting in 2000 training samples). Model parameter optimization uses the EM algorithm. The expectation step (E-step) calculates the posterior probability of the latent variables for each sample based on the current parameters (latent variables are set as potential influencing factors of each deformation state). The maximization step (M-step) updates the conditional probability table parameters of the model using gradient descent. The number of iterations is set to 100, and the convergence threshold is set to... The optimization stops when the parameter change is less than the convergence threshold after 5 consecutive iterations. Model validation uses 5-fold cross-validation, randomly dividing 2000 training samples into 5 non-overlapping subsets (400 samples per subset). One subset is selected as the validation set, and the remaining 4 subsets are used as the training set for 5 rounds of training and validation. The deformation prediction accuracy (requirement) is used to determine the model's performance. ) and the mean square error of the probability distribution (required) As a validation metric, the model's generalization ability is ensured. When the weighted input vector set is input into the pre-trained Bayesian network model for inference, the component preprocessing layer normalizes the component-related dimensional data (such as the weighted shaft diameter deviation and the coaxiality data of the large end hole) in the weighted input vector set (normalization range is [0,1]). Local feature extraction is performed using convolutional kernels of a certain size, and the output dimension is obtained after passing through the Sigmoid activation function. The component features include geometric deviation features, material uniformity features, and pre-processing quality features. The assembly execution layer, based on preset assembly process association rules (such as the association formula between the tooling positioning pin parameters and bolt preload parameters corresponding to the large end cap assembly process), maps and fuses the component features with the assembly tooling positioning parameters and assembly operation parameters in the weighted input vector set. By calculating the Euclidean distance (threshold set to 0.05mm) between the large end hole coaxiality feature and the tooling positioning pin position deviation in the component features, valid associated features are filtered, and the output dimension is [missing value]. The assembly feature vector is generated, with each element corresponding to the feature contribution value of each assembly process. The deformation generation layer, based on the component features and the assembly feature vector, can call a preset deformation probability inference formula. It first calculates the conditional mean and conditional variance of each node (corresponding to 6 deformation states), then calculates the posterior probability of each deformation state based on prior distribution parameters, outputting a deformation prediction probability distribution containing 6 deformation states and their corresponding probability values ​​(e.g., no deformation probability 0.72, slight deformation probability 0.21, mild deformation probability 0.05, other deformation state probability 0.02). The deformation prediction probability distribution is as follows: , in, For the deformation prediction probability distribution, For assembly deformation state variables, For component feature vectors, For assembly feature vectors, For the overall parameters of the Bayesian network model, The number of nodes in the deformation generation layer. For component feature weights, For assembly feature weights, For the corresponding number The component feature vectors of each node, For the corresponding number Assembly feature vectors of each node For the first The average value of the deformation state conditions of each node. For the first The conditional variance of the deformation state of each node. Let be the prior distribution parameters of the mean value of the deformation state conditions. Let be the prior distribution parameter of the variance of the deformation state condition. Let be the probability density function of a normal distribution. Let be the probability density function of the gamma distribution. For joint integral operators, Let be the probability density function of the conditional variance of the deformable state. This represents the conditional probability distribution of the assembly deformation state variables.

[0028] In some embodiments, step S103 involves statistical analysis of the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics. For a flowchart of the statistical analysis, please refer to [link to flowchart]. Figure 2 ,include: S1031. The deformation prediction probability distribution is sampled using a preset sliding window to obtain a probability distribution subsequence; Statistical methods including sliding window sampling, time series analysis, and function fitting are used to process the deformation prediction probability distribution output by the Bayesian network. The probability distribution is sampled through a preset window to obtain probability distribution subsequences. The preset sliding window duration can be set to 10 minutes (matching the average cycle of a single process in marine diesel engine connecting rod assembly, ensuring coverage of complete process data), and the step size can be set to 2 minutes (avoiding data redundancy and ensuring temporal continuity). Sliding sampling is performed on the real-time output deformation prediction probability distribution (recording one group every 2 minutes in chronological order, each group containing the probability values ​​of 6 deformation states). That is, starting from the first group of data, five consecutive groups of probability distribution data (corresponding to a 10-minute duration) are extracted each time as one probability distribution subsequence until all time series data are traversed, ultimately obtaining M probability distribution subsequences. ).

[0029] S1032. Calculate the mean of the probability distribution subsequences to construct a deformed probability time series sequence; The mean of the probability distribution subsequences is calculated to construct a time series sequence of deformation probabilities. For each probability distribution subsequence, the arithmetic mean of the probabilities corresponding to the six deformation states is calculated (e.g., in a certain subsequence, the five probability values ​​for "slight deformation" are 0.21, 0.23, 0.22, 0.24, and 0.22, with a mean of 0.224). The mean probabilities of each deformation state in all subsequences are arranged sequentially according to the sampling time sequence to form a sequence with dimension [missing information]. The deformation probability time series (each row corresponds to the mean of one window, and each column corresponds to one deformation state).

[0030] S1033. Perform trend fitting on the deformation probability time series to obtain the fitting function, and use the characteristic parameters of the fitting function as the deformation trend features. The mean of the subsequences is calculated to construct a deformation probability time series, and trend fitting is performed on the deformation probability time series to extract the feature parameters of the fitting function as trend features reflecting the overall direction of deformation change. For the probability time series data of the two key deformation states of "moderate deformation" and "severe deformation" (because these states have a significant impact on assembly quality), which are of the highest interest, the least squares method is used for trend fitting, and a quadratic polynomial function is selected. ) as the fitting function ( The sequence number is 1 to M. This represents the mean probability of deformation for the corresponding time series; , , The fitting coefficients are obtained by using the polyfit function in MATLAB (with the fitting order set to 2) to calculate the coefficient values, thus obtaining the fitting functions for the two key deformation states. The feature parameters of the fitting functions are extracted as deformation trend features, including the quadratic term coefficients. (Reflecting the trend curvature, This indicates that the probability increases convexly over time. (Indicates concave growth) coefficient of the linear term (Reflecting the slope of the trend, This indicates an overall upward trend. (Indicating an overall downward trend), constant term (Probability baseline value reflecting the initial time series) and the x-coordinate of the vertex of the fitted function (Reflecting the timing of trend reversals), a total of 4 characteristic parameters constitute the deformed trend feature set.

[0031] S1034. Calculate the residual between the deformation probability time series and the fitting function at each sampling point, calculate the confidence interval width of the deformation prediction probability distribution based on the residual, and calculate the information entropy of the deformation prediction probability distribution. Determine the distribution dispersion characteristics based on the confidence interval width and information entropy. By calculating the residuals of the time series and the fitted function at each sampling point, the uncertainty of the probability distribution is quantified. Based on the residuals, the confidence interval width and information entropy are calculated as distribution dispersion characteristics characterizing the volatility and reliability of the prediction results, providing a comprehensive quantitative basis for subsequent risk identification. The residuals of each sampling point in the deformed probability time series are calculated, i.e., for each time series number... Subtract the actual mean of the probability of "moderate deformation" or "severe deformation" corresponding to that sequence number from the value of the fitted function. The predicted value at each sampling point is used to obtain the residual, forming a residual sequence. The distribution dispersion characteristics are calculated based on the residual sequence, where the confidence interval width is calculated using... Confidence level: First, calculate the standard deviation using the residual series. The confidence interval width is calculated based on the standard deviation, and its confidence interval width is: , in, The width of the confidence interval; The information entropy of the probability values ​​of the six deformation states in the deformation prediction probability distribution is calculated using the Shannon entropy formula. The information entropy is: , in, For information entropy, For the first The probability value of each deformed state. The distribution dispersion feature is formed by the confidence interval width and the information entropy.

[0032] Key deformation risk factors are identified based on deformation trend characteristics and distribution dispersion characteristics. Please refer to the flowchart for the key deformation risk factor identification process. Figure 3 ,include: S1035. Calculate the correlation between multi-dimensional assembly data and deformation trend characteristics to obtain the first correlation value; The deformation trend features were expanded into four time series of features (each time series has the same length as the time series of the multi-dimensional assembly data, both being M). The Pearson correlation coefficient method was used to calculate the correlation coefficient between the time series of each dimension of assembly data (such as the coaxiality of the connecting rod big end hole, the position deviation of the tooling locating pin, and the bolt preload) and the time series of each set of deformation trend features. The maximum value among the four correlation coefficients was taken as the first correlation value between that dimension of data and the deformation trend (the value range is...). The absolute value is used to measure the strength of the association; the closer the absolute value is to 1, the stronger the association.

[0033] S1036. Calculate the correlation between multi-dimensional assembly data and distribution dispersion characteristics to obtain the second correlation value; The distribution dispersion features (confidence interval width, information entropy) are also expanded into two sets of feature time series in a time series manner. Using the same Pearson correlation coefficient method, the correlation coefficient between the time series of assembled data in each dimension and the time series of each set of distribution dispersion features is calculated. The maximum value of the two correlation coefficients is taken as the second correlation value between the data in that dimension and the distribution dispersion (the absolute value is also used to measure the correlation strength).

[0034] S1037. Based on the preset first correlation threshold and the first correlation value, the multi-dimensional assembly data is filtered to obtain the first type of candidate data; The determination of the preset threshold needs to be based on the statistical analysis of historical qualified assembly data. For example, select 100 qualified assembly cases with no deformation / slight deformation, calculate the first correlation value and the second correlation value of all dimensions of assembly data in each case, and take the 90th quantile of the first correlation value (i.e., The first correlation value of qualified cases is lower than this value. 0.65 is used as the preset first correlation threshold. Based on the first correlation threshold, the dimensional assembly data with a first correlation value (absolute value) of not less than 0.65 are selected, such as "coaxiality deviation of connecting rod big end hole" and "bolt preload fluctuation value", etc., and are recorded as the first type of candidate data.

[0035] S1038. Based on the preset second correlation threshold and the second correlation value, the multi-dimensional assembly data is filtered to obtain the second type of candidate data; The 90th percentile of the second correlation value, 0.6, is taken as the preset second correlation threshold. Based on the second correlation threshold, the assembly data of the dimension with a second correlation value (absolute value) of not less than 0.6, such as "deviation of tooling positioning pin" and "fluctuation of assembly environment temperature", are selected and recorded as the second type of candidate data.

[0036] S1039. Integrate the first type of candidate data and the second type of candidate data to obtain key assembly data, obtain the assembly process parameters corresponding to the key assembly data, and use the assembly process parameters as key deformation risk factors. Integrate the first and second categories of candidate data, remove duplicates (if a dimension of data meets both threshold conditions, only one instance is retained), and use the Grubbs test (significance level set to 0.05) to remove outliers in the integrated data (such as abnormally high correlation data caused by temporary sensor malfunctions) to form key assembly data, such as "connecting rod big end hole coaxiality deviation", "bolt preload fluctuation value", and "tooling locating pin position deviation". Extract the assembly process parameters corresponding to the key assembly data, such as: "connecting rod big end hole coaxiality deviation" corresponding to "tool wear threshold for big end hole finishing process ( "Coaxiality tolerance standard for the large-head hole inspection station ()" ")"; "Bolt preload fluctuation value" corresponds to "Torque wrench accuracy grade of bolt tightening process ()"; "Preload application procedure (applied in 3 steps, each step...") "Rated preload"; "Tooling positioning pin position deviation" corresponds to "positioning pin position detection frequency in tooling calibration process (calibrate once every 100 pieces)" and "positioning pin replacement cycle (replace after a cumulative use of 300 pieces)". These corresponding assembly process parameters are the key deformation risk factors.

[0037] In some embodiments, in step S104, a diesel engine connecting rod assembly deformation prediction report is generated based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors. The deformation prediction probability distribution obtained by Bayesian network inference, the deformation trend characteristics and distribution dispersion characteristics extracted by statistical analysis, and the key deformation risk factors identified through correlation analysis are integrated and formatted. A comprehensive assembly deformation prediction report containing deformation prediction results, risk quantification indicators, and process optimization suggestions is synthesized through a report generation template or structured document generation technology (such as automatically filling in preset fields, generating data visualization charts and risk level assessment conclusions), providing a direct basis for assembly process decision-making and optimization.

[0038] In summary, the diesel engine connecting rod assembly deformation prediction method provided by this invention acquires multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods, weights the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set, infers the weighted input vector set based on a pre-trained Bayesian network model to obtain a deformation prediction probability distribution, performs statistical analysis on the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics, identifies key deformation risk factors based on the deformation trend characteristics and distribution dispersion characteristics, and generates a diesel engine connecting rod assembly deformation prediction report based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors, thereby improving the assembly quality and product reliability of marine diesel engine connecting rods.

[0039] To better implement the diesel engine connecting rod assembly deformation prediction method in this embodiment of the invention, based on the diesel engine connecting rod assembly deformation prediction method, correspondingly, as follows: Figure 4 As shown, this embodiment of the invention also provides a diesel engine connecting rod assembly deformation prediction device. The diesel engine connecting rod assembly deformation prediction device 400 includes: Data acquisition module 401 is used to acquire multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods, and to weight the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set; The inference module 402 is used to infer the weighted input vector set based on the pre-trained Bayesian network model to obtain the deformation prediction probability distribution. The risk factor identification module 403 is used to perform statistical analysis on the deformation prediction probability distribution, obtain deformation trend characteristics and distribution dispersion characteristics, and identify key deformation risk factors based on the deformation trend characteristics and distribution dispersion characteristics. The prediction report generation module 404 is used to generate a diesel engine connecting rod assembly deformation prediction report based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors.

[0040] like Figure 5 As shown, the present invention also provides a diesel engine connecting rod assembly deformation prediction device 500, which can be a computing device such as a mobile terminal, desktop computer, laptop, handheld computer, or server. The diesel engine connecting rod assembly deformation prediction device 500 includes a processor 501, a memory 502, and a display 503. Figure 5 Only some components of the diesel engine connecting rod assembly deformation prediction device 500 are shown. However, it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0041] In some embodiments, the memory 502 can be an internal storage unit of the diesel engine connecting rod assembly deformation prediction device 500, such as a hard disk or memory of the diesel engine connecting rod assembly deformation prediction device 500. In other embodiments, the memory 502 can also be an external storage device of the diesel engine connecting rod assembly deformation prediction device 500, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the diesel engine connecting rod assembly deformation prediction device 500. Furthermore, the memory 502 can include both internal storage units and external storage devices of the diesel engine connecting rod assembly deformation prediction device 500. The memory 502 is used to store application software and various types of data installed on the diesel engine connecting rod assembly deformation prediction device 500, such as the program code installed on the diesel engine connecting rod assembly deformation prediction device 500. The memory 502 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 502 stores a diesel engine connecting rod assembly deformation prediction program, which can be executed by the processor 501 to realize the diesel engine connecting rod assembly deformation prediction method of various embodiments of the present invention.

[0042] In some embodiments, processor 501 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 502 or process data, such as a diesel engine connecting rod assembly deformation prediction method.

[0043] In some embodiments, display 503 may be an LED display, a liquid crystal display, a touch-screen liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 503 is used to display identification information from the diesel engine connecting rod assembly deformation prediction program and to display a visual user interface. Components 501-503 of the diesel engine connecting rod assembly deformation prediction device 500 communicate with each other via a system bus.

[0044] In some embodiments, when the processor 501 executes the diesel engine connecting rod assembly deformation prediction program in the memory 502, it implements each step of the diesel engine connecting rod assembly deformation prediction method as described in the above embodiments. Since the diesel engine connecting rod assembly deformation prediction method has been described in detail above, it will not be repeated here.

[0045] Accordingly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can realize the steps or functions in the diesel engine connecting rod assembly deformation prediction method provided in the above-described method embodiments.

[0046] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0047] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting deformation during diesel engine connecting rod assembly, characterized in that, include: Multi-dimensional assembly data during the assembly process of marine diesel engine connecting rod is obtained, and the multi-dimensional assembly data is weighted based on assembly process knowledge to obtain a weighted input vector set. The weighted input vector set is inferred based on a pre-trained Bayesian network model to obtain the deformation prediction probability distribution. Statistical analysis is performed on the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics, and key deformation risk factors are identified based on the deformation trend characteristics and distribution dispersion characteristics. Based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors, a deformation prediction report for diesel engine connecting rod assembly is generated.

2. The method for predicting deformation of diesel engine connecting rod assembly according to claim 1, characterized in that, The weighting of the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set includes: Obtain the deformation influencing factor and the quantitative value of the degree of influence of the deformation influencing factor in the connecting rod assembly process of the assembly process knowledge; Calculate the feature matching degree between the multi-dimensional assembly data and the deformation influencing factors; Calculate the weight value of each dimension of the assembly data based on the feature matching degree and the quantification value of the degree of influence. The assembly data for each dimension is weighted based on the weight values ​​to generate a weighted input vector set.

3. The method for predicting deformation of diesel engine connecting rod assembly according to claim 2, characterized in that, The Bayesian network model includes a component preprocessing layer, an assembly execution layer, and a deformation generation layer. The pre-trained Bayesian network model infers from the weighted input vector set to obtain the deformation prediction probability distribution, including: Based on the component preprocessing layer, feature extraction is performed on the weighted input vector set to obtain component features; Based on the assembly execution layer, the component features are assembled and mapped to obtain an assembly feature vector; Based on the deformation generation layer, deformation probability inference is performed on the component features and assembly feature vectors, and the deformation prediction probability distribution is output.

4. The method for predicting deformation of diesel engine connecting rod assembly according to claim 3, characterized in that, The deformation prediction probability distribution is as follows: ， in, For the probability distribution of deformation prediction, For assembly deformation state variables, For component feature vectors, For assembly feature vectors, For the overall parameters of the Bayesian network model, The number of nodes in the deformation generation layer. For component feature weights, For assembly feature weights, For the corresponding number The component feature vectors of each node, For the corresponding number Assembly feature vectors of each node For the first The average value of the deformation state conditions of each node. For the first The conditional variance of the deformation state of each node. Let be the prior distribution parameters of the mean value of the deformation state conditions. Let be the prior distribution parameter of the variance of the deformation state condition. Let be the probability density function of a normal distribution. Let be the probability density function of the gamma distribution. For joint integral operators, Let be the probability density function of the conditional variance of the deformable state. This represents the conditional probability distribution of the assembly deformation state variables.

5. The method for predicting deformation of diesel engine connecting rod assembly according to claim 3, characterized in that, The statistical analysis of the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics includes: A preset sliding window is used to sample the deformation prediction probability distribution to obtain a probability distribution subsequence; The mean of the probability distribution subsequences is calculated to construct a deformed probability time series sequence; A fitting function is obtained by performing trend fitting on the deformation probability time series, and the feature parameters of the fitting function are used as deformation trend features. Calculate the residual between the deformation probability time series and the fitting function at each sampling point, calculate the confidence interval width of the deformation prediction probability distribution based on the residual, calculate the information entropy of the deformation prediction probability distribution, and determine the distribution dispersion characteristics based on the confidence interval width and information entropy.

6. The method for predicting deformation of diesel engine connecting rod assembly according to claim 5, characterized in that, The identification of key deformation risk factors based on the deformation trend characteristics and distribution dispersion characteristics includes: Calculate the correlation between the multi-dimensional assembly data and the deformation trend characteristics to obtain a first correlation value; Calculate the correlation degree between the multi-dimensional assembly data and the distribution dispersion feature to obtain a second correlation degree value; The multi-dimensional assembly data is filtered based on a preset first correlation threshold and a first correlation value to obtain a first type of candidate data; The multi-dimensional assembly data is filtered based on a preset second correlation threshold and a second correlation value to obtain a second type of candidate data; The first type of candidate data and the second type of candidate data are integrated to obtain key assembly data, and the assembly process parameters corresponding to the key assembly data are obtained, and the assembly process parameters are used as key deformation risk factors.

7. The method for predicting deformation of diesel engine connecting rod assembly according to claim 6, characterized in that, The diesel engine connecting rod assembly deformation prediction report includes deformation prediction results, risk quantification indicators, and process optimization suggestions.

8. A diesel engine connecting rod assembly deformation prediction device, characterized in that, include: The data acquisition module is used to acquire multi-dimensional assembly data during the assembly process of marine diesel engine connecting rods, and to weight the multi-dimensional assembly data based on assembly process knowledge to obtain a weighted input vector set. The inference module is used to infer the weighted input vector set based on a pre-trained Bayesian network model to obtain the deformation prediction probability distribution. The risk factor identification module is used to perform statistical analysis on the deformation prediction probability distribution to obtain deformation trend characteristics and distribution dispersion characteristics, and to identify key deformation risk factors based on the deformation trend characteristics and distribution dispersion characteristics. The prediction report generation module is used to generate a diesel engine connecting rod assembly deformation prediction report based on the deformation prediction probability distribution, deformation trend characteristics, distribution dispersion characteristics, and key deformation risk factors.

9. A diesel engine connecting rod assembly deformation prediction device, characterized in that, Including memory and processor; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the diesel engine connecting rod assembly deformation prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the diesel engine connecting rod assembly deformation prediction method according to any one of claims 1-7.

Images