CN116777040A

CN116777040A - Multi-objective parameter optimization method for hard turning process

Info

Publication number: CN116777040A
Application number: CN202310456717.6A
Authority: CN
Inventors: 迟玉伦; 余建华; 范志辉; 黄浩伦; 顾开创; 张淑权; 高程远
Original assignee: University of Shanghai for Science and Technology
Current assignee: University of Shanghai for Science and Technology
Priority date: 2023-04-25
Filing date: 2023-04-25
Publication date: 2023-09-19

Abstract

The invention relates to a multi-objective parameter optimization method for a hard turning process, which aims at solving the problem of optimizing cutting parameters in the existing hard turning low-carbon manufacturing process and performs cooperative optimization on carbon emission and surface roughness of hard turning. Cutting power is an important influencing factor of the carbon emission in the hard turning process, and a model capable of accurately reflecting the carbon emission in the hard turning process is established by analyzing the cutting power through a test. And carrying out a surface roughness test under the hard turning cutting working condition, and carrying out multi-element nonlinear regression fitting on the cutting parameters and the surface roughness through an orthogonal test and a generalized regression neural network algorithm to obtain a prediction model capable of effectively predicting the surface roughness under the hard turning condition. And (3) converting the comprehensive optimization index into a single-target optimization model by using a linear weighting method, optimizing by using a genetic algorithm to obtain optimal cutting parameters, and finally verifying the validity of the multi-target optimization model in actual product processing of a factory. And the green manufacturing and process performance collaborative optimization of the hard turning is realized.

Description

Multi-objective parameter optimization method for hard turning process

Technical Field

The invention relates to a processing technology, in particular to a multi-objective parameter optimization method of a hard turning process for green high-surface quality manufacturing.

Background

During the manufacturing process of workpieces, the manufacturing industry generates a great amount of energy consumption and material consumption, and seriously affects the environment. Enterprises attach more importance to the environmental impact of product processing. How to comprehensively consider the carbon emission and the product quality in the processing process to optimize the cutting parameters is a hot problem to be solved in the manufacturing industry at present. Compared with the traditional finish machining method, namely grinding machining, the finish machining method for the hardened steel is higher in efficiency and economic benefit, can reduce environmental pollution, and meanwhile, the machined workpiece surface roughness can meet the machining precision requirement of the same grinding level. The surface roughness is an important index for measuring the integrity of the hard turning surface, and has obvious influence on the service performance, fatigue strength, contact stiffness, assembly precision and the like of a workpiece. If the cutting parameters are optimized with only low carbon emissions from the machining process in mind, the machining quality of the workpiece may be adversely affected. Therefore, the research on the optimization of the low-carbon low-surface roughness parameters of the hard turning is of great significance. The comprehensive consideration of reducing the carbon emission and the surface roughness to optimize the cutting parameters has important significance for energy conservation and emission reduction in manufacturing industry.

At present, a plurality of students at home and abroad research the optimization problem of the cutting parameters of the hard turning, finite element simulation is carried out on the cutting process of Cr12MoV of the cold work die steel of the dry-type hard turning of the opposite square boron nitride (CBN) tool such as Gao Shilong, the influence of the cutting parameters on the cutting force in the machining process is analyzed, and an empirical formula of the influence of the single cutting parameters on the cutting force is established. Minh et al studied the influence of various cutting parameters on the surface roughness by hard turning of cemented carbide with CBN tools, and the results showed that the feed rate had the greatest influence on the surface roughness under this condition. And (3) performing hard turning on AISI 52100 steel by using a PCBN tool by using UMAMAHESWARRAO and the like, and optimizing machining parameters by using a near ideal solution ordering method (TOPSIS) to obtain optimal cutting force, surface roughness and workpiece surface temperature. Most of the above studies are to optimize cutting parameters in terms of surface quality, cutting force, etc., and do not consider the influence of mechanical processing on the environment.

The parameter optimization model considering the environmental impact usually takes carbon emission, energy consumption or other environmental indexes as an optimization target, and Li Congbo and the like in early stages provide a process parameter optimization model taking minimum optimization time and minimum carbon emission as optimization targets, so that the consumption process of electric energy, cutting fluid and cutters is quantized into carbon emission, and the carbon emission is analyzed, and the optimization solution is carried out by using a complex-form method. Zhou Zhi identity takes the energy consumption and the machining efficiency of numerical control turning as optimization targets, and utilizes a multi-target teaching and learning algorithm to carry out optimization solution. Zhang Lei and the like model carbon emission and noise in the machining process under two working steps of turning the outer circle and threading, establish a low-carbon low-noise-oriented thread turning multi-objective optimization model, solve the multi-objective optimization model by utilizing a niche genetic algorithm, and the optimization result shows that the noise and the carbon emission are in a negative correlation. Fang et al studied the multi-step parameter optimization problem of milling, analyzed the carbon emission, production cost and processing time of the processing process, built a multi-objective optimization model with cutting parameters as variables, and provided an improved particle swarm optimization algorithm (PSO) for optimization. However, the researches on the above-mentioned several optimization models are focused on targets such as carbon emission, noise, processing time and the like in the manufacturing process of the workpiece, and on the aspect of processing product quality, only parameter constraints of theoretical processing quality are considered, and the optimization models are not built in combination with product quality indexes.

Experimental studies conducted by HELU M and the like also prove that if only the influence of the machining process on factors such as energy consumption, noise, production cost and the like is considered in the cutting parameter optimization problem, the necessary process performance of the workpiece can be sacrificed, and the quality problem of the workpiece is influenced, so that the green manufacturing problems such as cutting energy consumption, carbon emission and the like and the process performance are required to be comprehensively optimized, and effective theoretical support can be provided for actual machining. He Yan and the like propose a multi-objective optimization method which simultaneously considers the cutting specific energy, the surface roughness and the surface layer residual stress, and carry out collaborative optimization research on the process energy consumption and the process performance of the screw hard cyclone milling. Shailendra Pawanr and the like establish a turning surface roughness and energy consumption multi-objective optimization model for optimizing machining parameters in a turning process, and determine optimal turning parameters by adopting TOPSIS. The Jia S and the like introduce actual constraint conditions such as machine tool equipment performance, cutter service life and the like by analyzing the energy consumption characteristics of the machining process, and a multi-target optimization model which takes turning process parameters as optimization variables and takes low surface roughness and low energy consumption as an optimization target is established. Wang Qiulian and the like, a numerical control turning process parameter multi-objective optimization model is established, the influence of the process parameters on energy consumption, processing time and surface roughness is analyzed by using a response surface method, and an improved artificial bee colony algorithm is adopted to solve the optimal parameter combination. Feng et al propose a system method that combines an energy model, an experimental design, and a multi-objective optimization model simultaneously, and optimizes cutting parameters in consideration of energy consumption, processing time, and surface quality in milling. The above studies, while enabling multi-objective optimization of cutting energy consumption and process performance, have a number of drawbacks. Firstly, the cutting power in the machining process is affected by factors such as workpiece, cutter materials, cutting parameters and the like, and the cutting power is not clearly reflected in the research. Secondly, as the surface roughness of the workpiece is influenced by a plurality of factors such as cutting conditions, workpiece materials, cutter materials and the like, under different processing technologies, the influence of cutting parameters on the surface roughness is quite different, so if a prediction model can be built for the surface roughness obtained by a specific processing technology, the effectiveness of the multi-objective optimization model can be improved to the greatest extent.

Disclosure of Invention

Aiming at the problems that the current manufacturing industry increasingly pays attention to finish machining low energy consumption and low carbon emission, a multi-objective parameter optimization method of a hard turning process is provided, and the green manufacturing and process performance collaborative optimization of the hard turning process is realized.

The technical scheme of the invention is as follows: the multi-target parameter optimization method for the hard turning process specifically comprises the following steps:

1) Carrying out a surface roughness test under a hard turning cutting working condition, carrying out multiple nonlinear regression fitting on cutting parameters and surface roughness through an orthogonal test and a generalized regression neural network algorithm, and establishing a surface roughness prediction model;

2) Analyzing cutting power through a test, and establishing a carbon emission model reflecting the hard turning process;

3) Cutting three-element cutting speed V during turning _c A feed rate f and a back draft a _p The selection of the three elements directly affects carbon emission, is a main influencing factor of the surface roughness after processing, and selects three cutting elements as optimization variables;

4) On the basis of determining the optimization variables, the optimization indexes are converted into a single-target optimization model by using a linear weighting method comprehensive optimization index: because the dimension of the carbon emission function and the surface roughness function are different and the numerical value is greatly different, the two objective functions are normalized, the maximum value and the minimum value of the carbon emission function and the surface roughness function are respectively calculated, the maximum value and the minimum value are converted into a dimensionless number between 0 and 1 according to a formula (21),

Wherein CE represents a carbon emission function, and is obtained by converting the carbon emission model in the step 2) by taking an optimization variable as a function variable; VR represents a surface roughness function, and is obtained by converting the carbon emission model in the step 1) by taking an optimization variable as a function variable; CE (CE) ^* And VR (VR) ^* The dimensionless numbers of the carbon emission function and the surface roughness function after being processed are respectively represented;

the linear weighting method is utilized to convert the multi-objective function problem into a single-objective function problem, and the comprehensive optimization objective function of the surface roughness and the carbon emission is obtained, and the specific processing method is as follows:

in which W is ₁ And W is ₂ Weight coefficients respectively representing carbon emission and surface roughness are set according to actual conditions and related experiences for the weight coefficient values;

5) Optimizing the structural parameters of the surface roughness prediction model in the step 1) by using a genetic algorithm, and optimizing the parameters in the comprehensive optimization objective function in the step 4) by using the genetic algorithm to obtain a relatively optimal technological parameter combination;

6) And verifying the effectiveness of the comprehensive optimization objective function of the surface roughness and the carbon emission in the actual product processing of the factory, and realizing the rapid optimization of the cutting process parameters.

Further, the carbon emission in the hard turning process in the step 2) is divided into direct carbon emission and indirect carbon emission, wherein the direct carbon emission consists of no load, load and electric energy consumption in a tool changing state in the machine tool processing process; the indirect carbon emission is caused by material consumption, in the hard turning process, the loss of a cutter, cutting fluid, cut-off materials and raw materials is an important influencing factor of the carbon emission in the hard turning process, the allocation of the carbon emission of the cutter, the cutting fluid, the cut-off materials and the raw materials in the total use process of the preparation system is analyzed by taking the cutting time as a unit, the consumption of the raw materials is determined by process design, the treatment of the cut-off materials is carried out after the processing, and the carbon emission optimization force of the two parts is not considered;

The obtained carbon emission model in the hard turning process is as follows:

T _c indicating the replacement period of the cutting fluid; v (V) _f Representing the consumption of the cutting fluid;representing the concentration of the cutting fluid; f (f) _t Represents the carbon emission factor, m of the cutter _t Representing the mass of the tool; t is t _p Representing a preparation time; t is t _m Representing the cutting time; t is t _ct T represents the time for single tool change _t Indicating tool life considering the number of regrind times; p (P) _uo Is the lowest idle power; d represents the maximum diameter when the workpiece is refined; k (K) ₁ 、K ₂ Is the relative coefficient of the rotating speed of the main shaft of the machine tool; v (V) _c Representing the turning excircle speed; t (T) _c Indicating the replacement period of the cutting fluid; v (V) _f Representing the consumption of the cutting fluid; />Representing the concentration of the cutting fluid;

k. m, n, t represent coefficients related to cutting force, combined with cutting power P _c Equation (14) is determined by experimental fitting,

P _c ＝P _sp -P _u (14)

wherein P is _sp For spindle power, P, of machine tools during cutting _u Is no-load power.

Further, the surface roughness prediction model building method comprises the following steps: the cutting speed, the feeding amount and the back cutting amount which are set in the experiment are used as input layers of a generalized regression neural network, and the surface roughness measured after the workpiece is processed is used as an output layer; the input layer receives input signals, the number of neurons in the mode layer is equal to the number of samples in the training set, and nonlinear transformation is carried out on data from an input space to a mode space in the mode layer; there are two types of neurons in the summation layer: the first type of neuron calculates algebraic sum of all neurons of the pattern layer; the second type of neuron computes a weighted sum of mode layer neurons; finally, dividing two kinds of neurons of the summation layer by the output layer to obtain a predicted value of the surface roughness; in the whole operation process, the parameter smoothing factor sigma is optimized by adopting a genetic algorithm, the mean square error MSE of a neural network is used as an optimization target, global search is carried out on the smoothing factor in the value range, and the smoothing factor sigma is automatically matched to the smoothing factor value which is most suitable for the model.

Further, the parameters in the comprehensive optimization objective function in the step 4) are optimized through a genetic algorithm to obtain relatively optimal cutting parameter values, namely, under the constraint condition of a model, a group of optimal process combinations are searched in the cutting speed, the feeding amount and the back cutting amount, so that the comprehensive optimization objective function converted by the carbon emission function and the roughness model reaches the minimum value, and the method specifically comprises the following steps:

after the optimization variable, the objective function and the constraint condition are determined, the optimization variable is encoded, and the cutting speed V is encoded by adopting binary system _c A feed rate f and a back draft a _p Coding, in which the upper and lower limits of each element constituting an individual are used, i.e.

In the middle ofB _f 、/>Representing the binary encoded genetics;

b, a conversion method from a genetype to a variable:

according to V _c 、f、a _p The upper and lower limit values of the 3 variables are encoded by taking the genetics type as Gray codes to obtain the values of the variables, namely

And C, calculating the fitness of an objective function:

when crossing and mutation of genetic operators are carried out, if an individual which does not meet the constraint condition is generated, the individual needs to be eliminated, and the calculation formula of the fitness fit of the individual which meets the constraint condition is as follows:

Wherein F is ^U For the upper limit of the objective function F ^L A lower limit estimated value;

after the coding and the fitness calculation are completed, genetic algorithm solving can be performed.

The invention has the beneficial effects that: the multi-objective parameter optimization method of the hard turning process of the invention at least achieves the following effects,

1) Based on the hard turning machining principle, a multi-objective optimization model which takes the minimum surface roughness and the minimum carbon emission as optimization targets and takes the cutting speed, the feeding amount, the back cutting amount, the maximum cutting efficiency and the like in actual machining as constraints is established, and is converted into a single-objective (comprehensive optimization index) optimization model through a linear weighting sum method, and is optimized through a genetic algorithm.

2) In order to verify the effectiveness of the above-mentioned multi-objective optimization model, a hard turning experiment was performed on the actual bearing product of the factory for studying the carbon emissions and surface roughness obtained by optimizing the cutting parameters before and after. Designing a surface roughness orthogonal experiment taking cutting parameters as variables; the carbon emission model parameters such as the idle power parameter, the cutting power parameter and the like are obtained through the experiment, and a tamping foundation is laid for process optimization.

3) The experimental data are optimally analyzed based on the low-carbon low-surface roughness, and the result shows that: the comprehensive optimization index is the lowest when the cutting speed is 225m/min, the feeding amount is 0.08mm/r, and the back cutting amount is 0.10 mm. The carbon emission was increased by 13.05% but the surface roughness was reduced by 34.44% compared to before the optimization. The method is proved to be optimized by taking low carbon and low surface roughness as multiple targets, and the method can obtain the optimal technological parameter combination, thereby providing an effective solution for improving the product quality and controlling the carbon emission.

Drawings

FIG. 1a is a schematic view of an inner circular hard turning of a bearing ring;

FIG. 1b is a schematic diagram of hard turning;

FIG. 1c is a schematic illustration of hard car surface roughness formation;

FIG. 1d is a schematic representation of the microscopic shape of a hard vehicle surface;

FIG. 2 is a system boundary diagram of the present invention;

FIG. 3 is a block diagram of a GRNN network in accordance with the present invention;

FIG. 4 is a flowchart of a genetic algorithm in the method of the present invention;

FIG. 5 is a graph of the test idle power signal of the present invention;

FIG. 6 is a graph of the results of an idle power test of the present invention;

FIG. 7 is a graph of a test no-load power fit of the present invention;

FIG. 8a is a graph showing the effect of cutting speed and feed rate on carbon emissions in accordance with the present invention;

FIG. 8b is a graph showing the effect of back draft and cutting speed on carbon emissions in accordance with the present invention;

FIG. 8c is a graph showing the effect of back draft and feed on carbon emissions in accordance with the present invention;

FIG. 9 is a graph of the change in fitness function value (test set MSE) of the present invention;

FIG. 10 is a graph comparing training sets of GRNN surface roughness predictive models in accordance with the present invention;

FIG. 11 is a graph comparing test sets of GRNN surface roughness predictive models in accordance with the present invention;

FIG. 12a is a graph of fitness of the genetic algorithm of the present invention;

FIG. 12b is a combination of optimal cutting parameters obtained by optimizing the genetic algorithm of the present invention;

FIG. 13 is a graph of the effect of feed on objective functions obtained by the method of the present invention;

FIG. 14 a graph of the effect of cutting speed on various objective functions obtained by the method of the present invention;

FIG. 15A graph of the effect of back draft on various objective functions is obtained with the method of the present invention.

Detailed Description

The invention will now be described in detail with reference to the drawings and specific examples. The present embodiment is implemented on the premise of the technical scheme of the present invention, and a detailed implementation manner and a specific operation process are given, but the protection scope of the present invention is not limited to the following examples.

Aiming at the problem of optimizing cutting parameters of the existing hard turning low-carbon manufacturing, the method performs cooperative optimization of carbon emission and surface roughness on the hard turning. Cutting power is an important influencing factor of the carbon emission in the hard turning process, and a model capable of accurately reflecting the carbon emission in the hard turning process is established by analyzing the cutting power through a test. And carrying out a surface roughness test under the hard turning cutting working condition, and carrying out multi-element nonlinear regression fitting on the cutting parameters and the surface roughness through an orthogonal test and a generalized regression neural network algorithm to obtain a prediction model capable of effectively predicting the surface roughness under the hard turning condition. And (3) converting the comprehensive optimization index into a single-target optimization model by using a linear weighting method, optimizing by using a genetic algorithm to obtain optimal cutting parameters, and finally verifying the validity of the multi-target optimization model in actual product processing of a factory.

1. Hard turning problem description and optimization variable determination:

the hard turning mode of the turning mill can ensure good product quality and reduce environmental pollution, and is gradually applied to mass product production, if the hard turning quality is ensured, certain carbon emission is reduced, and the method has important significance to the hard turning industry.

Taking hard turning of a bearing as an example, on the premise of ensuring the quality of a product, the influence of the machining process on the environment is comprehensively considered. The raceways of the inner and outer rings are the main failure parts of the bearing in the use process, and the surface roughness of the raceways of the rings can influence the surface lubrication state and the friction factor. The surface roughness is an important index for judging the processing quality of the bearing, and under a complex working condition, the surface quality has a great influence on the service performance of the bearing, and serious abrasion can cause the bearing to fail very quickly, so that the normal service life and the durability of the bearing are affected. Therefore, it is studied to reduce the surface roughness of the bearing raceway from the viewpoint of ensuring the machining quality, and the hard turning surface roughness is schematically shown in fig. 1a to 1 d.

The system boundaries are shown in fig. 2: and establishing a multi-objective optimization model by taking the low carbon emission and the low surface roughness as optimization targets. The carbon emission in the hard turning process is mainly divided into direct carbon emission and indirect carbon emission, and the direct carbon emission mainly comprises no-load, load and electric energy consumption in a tool changing state in the machine tool processing process. The indirect carbon emission is caused by material consumption, in the hard turning process, the loss of a cutter, cutting fluid, cut-off materials and raw materials is an important influencing factor of the carbon emission in the hard turning process, the carbon emission amount of the cutter, the cutting fluid, the cut-off materials and the raw materials in the preparation system is analyzed to be apportioned in the total using process of the cutter, the cutting fluid, the cut-off materials and the raw materials in the unit of cutting time, the consumption of the raw materials is determined by process design, the treatment of the cut-off materials (scraps) is carried out after the processing, and therefore, the optimizing force of the carbon emission amount of the two parts of the consumption of the raw materials and the treatment of the cut-off materials is limited and is not considered for the moment. The surface roughness (shown in figure 1 c) is an important index reflecting the surface quality and the error of the microscopic geometry (shown in figure 1 d) of the part, and is an important criterion for measuring the quality of a hard turning bearing, and in the hard turning process, the cutting speed V is commonly controlled as shown in figures 1b and 1c _c A feeding amount f and a back cutting amount a _p To optimize the surface roughness of the workpiece. In the process of optimizing technological parameters, orthogonal experiments are designed, a surface roughness prediction model is established based on experimental data, and relatively optimal cutting parameters are sought. By comprehensively considering two optimization targets of low carbon emission and low surface roughness, the cutting parameters which can improve the product quality and reduce the environmental pollution can be obtained.

Cutting speed V during turning _c A feed rate f and a back draft a _p The cutting time determined by the three factors is used for influencing the material consumption of a cutter and cutting fluid, the three factors are main influencing factors of the surface roughness after machining, and the change of each factor can influence the carbon emission and the surface roughness, so the three factors are selected as optimization variables.

2. Establishing a mathematical model of carbon emission in the hard turning process:

study of carbon emission process of finish turning raceway of bearing hard turning, according to the analysis, carbon emission C of hard turning process _p Comprising 3 parts: carbon emission C of turning consumed electric energy _e Carbon emission C of cutter loss _t And cutting fluid consumption carbon emission C _c . Thus, the carbon emission function model of the hard turning process is shown in formula (1):

C _p ＝C _e +C _t +C _c (1)

2.1 hard turning consuming electric energy carbon emission C _e And (3) resolving:

in actual production, the processing time t of one procedure is set to be the preparation time t _p Cutting time t _m And average tool change time t after tool wear _c The mathematical model of the composition, processing time function is as follows:

wherein t is _ct T represents the time for single tool change _t The tool life is expressed taking the number of regrind into account.

The energy consumption of the hard turning process comprises the energy consumption of the idle state, the load state and the tool changing state. No-load power P _u Indicating the power of a machine tool when idling, additional load loss power P is generated when a machine tool is switched from idling to a load state _a Cutting power during turning is P _c The spindle stops rotating when the tool is changed, and the power at the moment is the tool changing power P _e . The energy approximate balance equation during the dynamic operation of the machine tool is as follows:

when the main shaft of the machine tool stably runs at a fixed rotating speed and the load is fixed, the fluctuation of the total input power, the no-load power, the load power and the load loss power is small, and the fluctuation can be considered as a constant value, and the total energy consumption can be expressed as:

the hard turning process consumes the following electric energy and carbon emissions:

C _e ＝F _e ·E _e (5)

wherein F is _e Representing the carbon emission factor (kgCO) of the consumption of electric energy ₂ /kWh)，E _e Representing the electrical energy consumed by the hard turning process.

Hard turning time stage calculation:

(1) Turning excircle time t _m Solution calculation

Wherein L represents the processing length of the workpiece, D represents the maximum diameter of the workpiece during polishing, y _r +Δ represents the in-cut and over-cut that the actual cutting length needs to take into account.

Turning excircle speed:

in the formula, w represents the spindle rotation speed during the motor cycle.

(2) Tool changing time t _c Solution calculation

Tool durability:

wherein C is _T The constants related to the cutting conditions are expressed, and x, y, and z represent coefficients related to the tool life.

Thus, the tool change time:

and (3) calculating the power in the processing process:

(1) No-load power P _u

In the process of hard turning, when the machine tool is in idle load, the idle load power of the machine tool has great influence on the power loss of the machine tool, and the composition of the loss power of the machine tool can be accurately reflected only when the idle load power of the machine tool and the rotating speed of the main shaft approximately have a quadratic function change relation, and the idle load power P _u The quadratic function relation with the rotating speed of the main shaft of the machine tool is as follows:

P _u ＝P _uo +K ₁ w+K ₂ w ² (10)

wherein K is ₁ 、K ₂ Is the relative coefficient of the rotating speed of the main shaft of the machine tool, P _uo Is the lowest idle power.

(2) Tool changing power P _e

The main shaft of the machine tool stops rotating when the tool is changed, and the power of the machine tool is the lowest idle power, so that the tool changing power is as follows:

P _e ＝P _uo (11)

(3) Cutting power P _c

P _c ＝F _c V _c (12)

Wherein F is _c Representing the cutting force.

According to the empirical formula of the cutting force, an exponential relation formula between the cutting power and the cutting amount can be further deduced:

wherein k, m, n, t represents a coefficient related to the cutting force, and the cutting power P can be combined _c Equation (14) is determined by experimental fitting.

P _c ＝P _sp -P _u (14)

Wherein P is _sp The spindle power of the machine tool during cutting.

(4) Additional load power P _a

Parasitic load loss power P _a The load loss is approximately in linear proportion relation, and the expression is as follows:

P _a ＝b _m P _c (15)

wherein b is _m The load loss factor is usually 0.15 to 0.25 in actual processing.

2.2 cutter Egrew carbon emission solution

In actual hard turning, the direct carbon emissions caused by the tool are very small and may not be considered, so only indirect carbon emissions due to tool loss, which refers to the apportionment of the carbon emissions generated by the tool used in the cutting process during its manufacture, during the use of the tool, are considered herein. In the actual machining process, the cutter possibly comprises multiple regrinding, and the service life of the cutter is as follows:

T _t ＝(N+1)T (16)

in the formula, N represents the number of regrind times.

Therefore, the cutter loss carbon emission is as follows:

wherein f _t Indicating knifeHaving a carbon emission factor, m _t Indicating the mass of the tool.

2.3, loss of carbon emission from cutting fluids

The hard turning is usually a dry cutting method without cutting fluid, but the experimental object is bearing hard turning, and certain requirements are imposed on the surface quality of a workpiece and the service life of a cutter, so that a water-based cutting fluid is used in machining, and a continuous and uniform cooling mode (the cutting fluid comprises a water-based cutting fluid and an oil-based cutting fluid) is adopted, so that microcracks of the blade are avoided. The amount of carbon emissions caused by the cutting fluid is considered from both the amount of carbon emissions of the production cutting fluid and the amount of carbon emissions of the treatment cutting waste fluid. Therefore, the carbon emission factor of the cutting fluid is divided into the production of the carbon emission factor F of the cutting fluid _pe And treating the carbon emission factor F of the cutting fluid _me . Since the waste cutting fluid has a low mineral oil content, the waste cutting fluid treated carbon emission factor can be replaced with the waste water treated carbon emission factor.

The carbon emission loss of the cutting fluid is as follows:

wherein T is _c Indicating the replacement period of cutting fluid, V _f The amount of the cutting fluid is expressed,indicating the cutting fluid concentration.

Through the analysis, the mathematical model of the carbon emission in the hard turning process is as follows:

3. establishment of hard turning workpiece surface roughness model

After the workpiece is turned in a hard state, the surface roughness is an important index for evaluating the processing quality of the workpiece. Taking the influence of cutting parameters on the surface roughness of a workpiece as a way, combining orthogonal experimental data, optimizing GRNN structural parameters through a genetic algorithm based on a Generalized Regression Neural Network (GRNN) algorithm, and establishing a prediction model of the relation between machining parameters (cutting speed, feeding amount and back cutting amount) and roughness numerical values, wherein the prediction model is shown in a formula (20).

VR＝f(V _c ,f,a _p ) (20)

FIG. 3 shows a network structure of a GRNN surface roughness prediction model, and compared with other neural network algorithms, the GRNN of the generalized regression neural network has excellent nonlinear mapping capability on the function point fitting approximation problem, and can achieve a good prediction effect when the problems of small sample data amount and unstable data are processed. The GRNN network structure has 4 layers, an input layer, a mode layer, a summing layer, and an output layer, respectively. And taking the experimentally set cutting speed, feeding amount and back cutting amount as input layers of the GRNN neural network, and taking the surface roughness measured after workpiece processing as output layers. The input layer receives the input signal, and the number of neurons in the pattern layer is generally equal to the number of samples in the training set, at which layer the data from input space to pattern space is transformed non-linearly. There are two types of neurons in the summation layer: the first type of neuron calculates algebraic sum of all neurons of the pattern layer; the second type of neuron computes a weighted sum of the mode layer neurons. And finally, dividing the two neurons of the summation layer by the output layer to obtain the predicted value of the surface roughness. In the whole operation process, the parameter sigma (smoothing factor) is a key for improving the prediction performance of the GRNN model. For the value of the smoothing factor, genetic algorithm optimization is selected to replace the traditional manual numerical value adjusting method, the mean square error MSE of the neural network is used as an optimization target, global search is carried out on the smoothing factor in the value range, and the smoothing factor value which is most suitable for the model can be automatically matched only by properly changing the parameter setting of the genetic algorithm.

4. Hard turning multi-objective optimization model solving method based on genetic algorithm

4.1 Multi-objective function transformation

Relates to two optimization targets of low carbon emission and low roughness, and belongs to the nonlinear optimization problem of multi-target and multi-decision variables. In dealing with the multi-objective optimization problem, since the objective functions are not simply linear, and the objective functions may collide with each other, the optimal values of the objective functions cannot be obtained at the same time. And a linear weighting method is selected to convert the multi-objective function into a single-objective function for processing. Since the carbon emission function and the surface roughness function have different dimensions and have great differences in numerical values, normalization processing is required for the two objective functions, and the maximum value and the minimum value of the carbon emission function and the surface roughness function are respectively calculated and converted into a dimensionless number between 0 and 1 according to the following formula.

Wherein CE represents a carbon emission function, VR represents a surface roughness function, CE ^* And VR (VR) ^* And respectively representing dimensionless numbers of the carbon emission function and the surface roughness function after being processed.

The specific processing method after converting the multi-objective function problem into the single-objective function problem (namely, the comprehensive optimization index) by using the linear weighting method comprises the following steps:

In which W is ₁ And W is ₂ Weight coefficients respectively representing carbon emission and surface roughness are set according to actual conditions and related experiences, and the optimal weight coefficient is obtained according to multiple experiments and statistics of actual production and processing of factories and is W ₁ ＝0.35，W ₂ ＝0.65。

4.2 model constraints

In the actual machining process, the value of the cutting parameter is influenced by a plurality of factors, and the limitation of conditions such as cutting speed, feeding amount, back cutting amount, workpiece machining quality, maximum cutting power, maximum cutting force and the like in cutting is considered, so that the cutting parameter is optimized in a numerical range meeting the limiting conditions.

1) Cutting speed constraint: the cutting speed during machining is limited by the spindle speed of the machine tool, which must be between the highest and lowest speeds of the machine tool, and taking into account the actual machining parameter requirements of the hard turning finish.

Wherein w is _min Represents the minimum rotation speed of the main shaft, w _max Indicating the maximum rotational speed of the spindle.

2) And (3) feed quantity constraint: the feeding amount is required to be in the range of the minimum feeding amount and the maximum feeding amount of the machine tool, and meets the actual requirement of hard turning finish machining.

f _min ≤f≤f _max (24)

Wherein f _min Representing the minimum feed rate of hard turning finish machining, f _max Indicating the maximum feed rate of the hard turning finish.

3) Back draft constraint: the back cutting amount is selected according to the factors of a processing technology, a workpiece material, a cutter material and the like, the surface quality of the workpiece is required to be high by a researched object, and the back cutting amount is considered to be a value in a hard turning finish machining range.

a _{p min} ≤a _p ≤a _{p max} (25)

Wherein a is _{p min} Represents the minimum back cutting tool amount, a, of hard turning finish _{p max} Indicating the maximum back draft of the hard turning finish.

4) Cutting power constraint: the cutting power should not be greater than the maximum cutting power prescribed by the machine tool, i.e

P _c ·η≤P _max (26)

Wherein eta represents the effective coefficient of machine tool machining power and P _max Indicating the maximum available cutting power specified by the machine tool.

4.3 solving the objective function based on the genetic algorithm

Based on the genetic algorithm, the transformed objective function (equation (22)) is optimized to obtain a relatively optimal value of the cutting parameter. The formulae (23) to (26) are constraints of the model. The optimal technological combination is found in the cutting speed, feeding amount and back cutting amount to make the carbon exhaust function and roughness model converted function reach minimum.

4.3.1 individual representations

After determining the optimization variables, objective functions and constraints, the optimization variables need to be encoded, here the cutting speed V using binary encoding _c A feed rate f and a back draft a _p Coding, in which the upper and lower limits of each element constituting an individual are used, i.e.

In the middle ofB _f 、/>Representing the binary encoded genetics.

4.3.2 conversion method from genotype to variable

According to V _c 、f、a _p The upper and lower limit values of the 3 variables are obtained by encoding the genetic type as Gray code, namely

4.3.3 fitness calculation of the objective function

When crossing and mutation of genetic operators are performed, if an individual which does not meet the constraint condition is generated, the individual needs to be eliminated. For an individual meeting the constraint condition, the calculation formula of the fitness fit is as follows:

wherein F is ^U For the upper limit of the objective function F ^L Is a lower limit estimate.

After the coding and the fitness calculation are completed, the genetic algorithm can be solved, and the operation flow of the genetic algorithm is shown in fig. 4.

5. Hard turning experimental setup

In order to verify the effectiveness of the model, a surface roughness test with three cutting elements as variables is carried out in the actual hard bearing product application of a factory, and parameters of related formulas such as idle power and the like are determined according to experimental conditions, so that the process parameters of the hard bearing product are optimized.

5.1, test conditions and protocol

The validity of the model is verified through a hard turning test. The test selects an RFCX26 horizontal numerical control lathe, the power of a main shaft of the machine tool is 15kw, the highest rotating speed can reach 4500r/min, and the related specification parameters of the machine tool are shown in a table 1; and processing the rolling bearing, wherein the workpiece is GCr15 bearing quenched steel, carrying out heat treatment at 1040 ℃ and quenching, and then tempering at a certain temperature to obtain a batch of test pieces with average hardness of about HRC60, wherein the maximum diameter of the workpiece is 75mm, and the maximum axial length of the workpiece is 13.6mm. And clamping the substitute machining workpiece on a workpiece shaft of the machine tool. The number of the cutters in actual use in the experiment is five, namely a first cutter is used for boring, a second cutter is used for turning a sealing groove, a third cutter is used for finely turning the bottom surface of a raceway, a fourth cutter is used for roughly turning the raceway, and a fifth cutter is used for finely turning the raceway; wherein, the finish turning raceway process determines the surface roughness of the bearing raceway, PCBN tools are selected for hard turning, the CBN content of the tools is 50%, the arc radius of the tool nose is 0.8mm, and other tool geometric parameters are shown in table 1.

TABLE 1

In order to study the influence of cutting parameters on the surface roughness of a workpiece in hard turning, the test determines an optimal technological parameter combination by taking the cutting parameters of a bearing finish turning raceway stage as factors through an orthogonal experiment. In the orthogonal experimental design, 5 levels of each factor were used, and an orthogonal test was performed at a level of 3 factors and 5, and the tested factor levels are shown in table 2. The surface roughness of the bearing raceway was measured using a roughness probe after each turning was completed. 2 points are selected on the surface of the workpiece during measurement, and the arithmetic average value of the measured value Ra of the points is used as experimental data; the test parameter settings and measurement results are shown in table 3. Meanwhile, in order to fit the cutting power in the carbon emission mathematical model, the cutting power during the machining of each set of test workpieces was monitored. Real-time monitoring of machine tool power is achieved through a De-tech U2044XA power sensor. The equipment is arranged at a main power supply of an electric cabinet machine tool to acquire total current and total voltage, a main transmission system current is acquired at a main shaft servo system to acquire machine tool power, a high-speed acquisition card of an NI company is adopted for data acquisition in a test, labview is used for writing data acquisition software, and the channel sampling frequency is set to 2000Hz.

TABLE 2

TABLE 3 Table 3

In addition, no-load power P is performed _u Fitting experiment and cutting Power P _c Fitting experiments. To study the idle power P of the machine tool _u And acquiring idle power data of the machine tool through a test according to the relation between the idle power data and the spindle rotating speed w, and performing secondary fitting by using MATLAB.16 groups of tests were carried out at a rotational speed of 500r/min to 2000/min and a step length of 100, and the test results are shown in Table 4 as idle power test data.

TABLE 4 Table 4

5.2 model basic parameter settings

1) Tool life, cutting fluid and other parameters:

parameters related to calculation of carbon emission functions such as cutter life, cutting fluid carbon emission factors and the like are set as follows: the test is a finishing stage for machining a bearing raceway, and the machining surface roughness R _amax It is required that the particle diameter should not exceed 0.5. Mu.m. Tool life-related parameters As shown in Table 5, when a cubic boron nitride tool is used for machining, the workpiece material is Gcr15 bearing steel, and the machining conditions are a precision, the cutting amount is three elements and the Taylor empirical formula parameter C of the tool durability _T X, y and z are 152.9, 0.52, 0.25 and 0.03 respectively.

TABLE 5

Obtaining carbon emission factor F of the east China area where the experiment is located by inquiring the reference line emission factor of the power grid of the China area _e 0.7921. Studies by RAJEMI et al showed that the average mass of individual blades was 9.5g, considering only the energy consumption of the tool preparation process, the total energy consumption of the tool was 1.5MJ. And combined with the electric energy carbon emission factor P _e Calculating the carbon emission factor f of the cutter _t 34.7kg CO ₂ /kg. Other carbon emission function related parameter settings are shown in Table 6, where t _ct 、t _p 、 _Tc 、V _f 、m _t N is determined according to the actual processing working conditions; load loss factor b _m According to reference (Zhou Zhiheng, zhang Chaoyong, xie Yang, huang Zhengtao, shao Xinyu. Energy efficiency of cutting parameters of numerically controlled lathesRate optimization [ J]Computer integrated manufacturing system, 2015,21 (09): 2410-2418) settings; cutting fluid carbon emission factor F _pe And treating the carbon emission factor F of the cutting waste liquid _me Low-carbon low-noise-oriented thread turning process parameter optimization [ J ] by reference (Zhang Lei, zhang Bei, bao Hong)]Computer integrated manufacturing system, 2018,24 (03): 639-648); the cutting amount and the excessive cutting amount y+delta are set by reference (Ai Xing, shoshi class. Handbook of cutting amount conciseness [ M)]Third edition Beijing: mechanical industry press, 2000).

TABLE 6

2) Idle power parameters:

determination of the minimum No. P in equation (10) by No. Power fitting experiments _uo Coefficient K related to rotation speed of main shaft of machine tool ₁ 、K ₂ . FIG. 5 shows the machine tool power signal at spindle speeds of 1200-1500r/min collected from the test. It can be seen in the signal diagram: at the moment after the rotating speed of the main shaft is increased, the power signal value of the machine tool has obvious sharp increase, then the machine tool immediately tends to be stable, and the power at the moment is no-load power at the current rotating speed of the machine tool; after power is collected for a period of time, the spindle is stopped to rotate, so that the power signal of the machine tool is rapidly reduced and then tends to be stable, and the power signal at the moment is the standby power of the machine tool. As can be seen from the graph, the average value of the oscillation range of the power signal shows an ascending trend in the process of increasing the rotation speed of the main shaft.

FIG. 6 shows the idle power test results of the spindle speed of the machine tool from 500r/min to 2000 r/min. From the experimental results, the idle power is basically in linear relation with the spindle rotation speed. Fitting the quadratic polynomial by using a least squares method to obtain the lowest no-load power P in the formula (10) _uo 250.1W, rotation speed correlation coefficient K ₁ And K ₂ 0.4759, -1.399X10 respectively ^-6 The fitting result is shown in fig. 7. The fitting goodness test is carried out on the regression model, wherein the R-square=99.78%, and the larger the value is, the better the regression model is fitted with the data; R-Sq (adj) =99.75%, which is equal to R-SquareThe degree of approach of (2) indicates the reliability of the regression model, and the test result shows that the degree of fit of the no-load power function is good. Meanwhile, the lowest idle power when the rotating speed is zero is very close to the standby power in test data, the accuracy of the quadratic polynomial fitting is verified, and the relation between the rotating speed of the main shaft and the idle power under the test condition can be effectively predicted.

3) Cutting power parameters:

the relation between cutting power and three cutting elements is complex, and the cutting power data obtained through experiments are used for fitting. Transforming equation (14) and taking the logarithm of it on both sides to obtain equation (34):

lgP _c ＝lgk+mlga _p +nlgV _c +tlgf (34)

lgP in equation (34) _c ，lgk，lga _p ，lgV _c ，l _g f is Y, a respectively ₀ ，x ₁ ，x ₂ ，x ₃ Representation, resulting in simplified formula (35):

Y＝a ₀ +mx ₁ +nx ₂ +tx ₃ (35)

and (3) performing multiple linear fitting operation on the formula (35), obtaining regression coefficients, calculating specific numerical values by utilizing MATLAB, and converting the formula (35) into an exponential form of cutting power to obtain k, m, n and t which are 2.6792,0.5153,1.1487,0.2499 respectively. In order to test the fitting degree of the cutting power model and the experimental value, the fitting degree is tested by R-square= 98.40%, the fitting degree is good, and the cutting power model can be matched with the cutting power in actual machining well.

6. Analysis of test results

And optimizing and analyzing the experimental data by using the hard turning optimization model. Firstly, establishing a mathematical model of the carbon emission amount in the processing process according to the parameters set by the test. And secondly, a surface roughness prediction model is built based on GRNN, structural parameters of the prediction model are optimized by GA, and model prediction accuracy is improved. And optimizing the multi-objective optimization model through a genetic algorithm to obtain a relatively optimal technological parameter combination. And finally, analyzing the influence of three cutting elements on the surface roughness, the carbon emission and the comprehensive optimization index, realizing the rapid optimization of the cutting process parameters, and improving the production benefits of enterprises.

6.1, carbon emission mathematical model

According to the experimental setting, the obtained no-load power coefficient K ₁ Cutting power coefficient k, tool life correlation coefficient C _T Substituting each parameter into a carbon emission mathematical model type (19) to obtain a mathematical model of the carbon emission in the processing process, wherein the mathematical model is as follows:

substituting the formula (36) into the formula (21) to obtain a carbon emission function CE, and obtaining the machining process carbon emission value under different cutting parameters as a carbon emission model part in the comprehensive optimization target formula (22). Analysis formula (36) shows that the carbon emission of the machining process is influenced by each cutting parameter, and in order to study the influence of each cutting parameter on the carbon emission of the machining process, a curve graph formed by interaction between two factors is drawn by using a Matlab software contour function, so that the coupling relation between every two factors can be obtained. The coupling relation of the factors is represented by the change trend of the influence of the two factors on the carbon emission function at the same time, and as shown in fig. 8a, 8b and 8c, each two cutting parameters form a trend graph of the influence on the carbon emission. As shown in fig. 8a, the effect of cutting speed and feed amount on carbon emission, when the feed amount is within [0.02,0.09] mm/r, the carbon emission amount decreases as the cutting speed increases; when the feed amount exceeds 0.09mm/r, the carbon emission amount is decreased and then increased as the cutting speed increases. As shown in fig. 8b, the influence of the cutting speed and the back draft on the carbon emission, when the back draft is [0.06,0.20], the carbon emission gradually decreases as the cutting speed increases; when the back cutting rate exceeds 170m/min in the range of 0.02,0.06 mm, the carbon emission is reduced and increased with the increase of the cutting rate. As shown in fig. 8c, the effect of the feed amount and the back draft on the carbon emission, when the back draft is unchanged, the carbon emission gradually increases as the feed amount increases; when the feed amount is unchanged, the increase in the back draft has little effect on the carbon emission amount.

The coupling relation of the two factors can show that the synergistic effect of the cutting speed, the feeding amount and the back cutting amount has an influence on the carbon emission amount. Wherein the synergistic effect of the cutting speed and the feeding amount has the greatest influence on the carbon emission amount, and the synergistic effect of the back cutting amount and the feeding amount has the smallest influence on the carbon emission amount. In order to study the influence of each cutting parameter on the surface roughness, an objective function of the surface roughness needs to be established, so that the surface roughness prediction of the technological parameters of cutting speed, feeding amount and back cutting amount is realized.

6.2 GRNN-based surface roughness prediction model

According to the surface roughness result of the table 3 obtained by the above experiment, a GRNN surface roughness prediction model type (21) is established, 80% of data (20 groups) are randomly selected as a training set of the model, and 20% (5 groups) of data are left as a test set of the model. The population size of the smoothing factor is set to be 50, the evolution algebra is set to be 10, the ratio of the crossed offspring is set to be 0.60, the searching range of the smoothing factor is set to be 0,1, the mean square error MSE is set as the fitness function, the obtained fitness change curve is shown in fig. 9, it can be seen that the smoothing factor is basically unchanged when the smoothing factor is evolved to the 5 th generation, and the searched smoothing factor is set to be 0.03.

And substituting the smooth factors obtained through GA optimization into a GRNN model, and comparing the smooth factors with a common GRNN model. As shown in FIG. 10, the average error between the predicted value and the true value of the training set of the GRNN model is 0.105, and the average error between the predicted value and the true value of the training set of the GA-GRNN model is 0.060. As shown in FIG. 11, the average error between the predicted value and the true value of the test set of the GRNN model was 0.151, and the average error between the predicted value and the true value of the test set of the GA-GRNN model was 0.054. The prediction results of the training set and the test set can show that the prediction effect of the GA-GRNN model is more accurate compared with the common GRNN model.

After the analysis, GA-GRNN with more accurate prediction effect is selected for predicting the surface roughness. Substituting the established surface roughness prediction model and the carbon emission function formula (36) into the comprehensive optimization objective function formula (22), and carrying out optimization solution on the objective function by utilizing a genetic algorithm to obtain an optimal process parameter combination.

6.3 model optimization based on genetic Algorithm

And calculating the established GRNN model by using a network simulation function (sim) in MATLAB, and forming a comprehensive optimization index with a carbon emission mathematical model as shown in a formula (22), namely establishing a nonlinear regression model taking low carbon emission and low surface roughness as optimization targets and taking cutting speed, feeding amount and back cutting amount as variables. And (3) solving and optimizing by using a GADST tool box, setting the population size to be 80, stopping algebra to be 50, elite number to be 10, and intersecting offspring proportion to be 0.70, so as to calculate the fitness function value and the optimal individual. When the optimization is completed, a population average fitness function value and an optimal individual fitness function value change curve are obtained, as shown in fig. 12a, and the fitness function value of the optimal individual gradually converges and tends to be stable along with the increase of population algebra. The optimal individual obtained by the optimization is shown in fig. 12 b: (V) _c ，f，a _p )＝(225，0.08，0.10)。

The experimental optimization results are shown in Table 7, and the surface roughness is maximum and 0.535 μm although the carbon emission is minimum and 165.09g when the low carbon emission is used as the optimization target; when the surface roughness is taken as an optimization target, the surface roughness value is minimum and is 0.104 mu m, but the carbon emission is maximum and is 630.53g; when the comprehensive index is used as an optimization target, the carbon emission is increased to 227.24g compared with 201g of the carbon emission of the parameter before optimization, and is increased by 13.05%, but the surface roughness value is greatly reduced to 0.198 μm compared with 0.302 μm of the surface roughness of the parameter before optimization, and is reduced by 34.44%. Therefore, the best comprehensive optimization value can be obtained with the comprehensive index as the optimization target.

TABLE 7

/>

After the optimal technological parameters optimized by the comprehensive indexes are combined, the green performance of the manufacturing process and the quality of the workpiece product are balanced. In order to study the influence degree of each cutting parameter on the comprehensive optimization target, single factor analysis is needed to be carried out on each cutting parameter, so that a cutting parameter range with acceptable comprehensive indexes is obtained, and the machining parameters can be flexibly adjusted by enterprises according to the influence factors such as production efficiency, machining energy consumption, manufacturing cost and the like during actual production and manufacturing.

6.4 influence of cutting parameters on the respective objective functions

In order to study the influence of each cutting parameter on the comprehensive optimization target and facilitate the flexible optimization of the cutting parameters, the influence of each cutting parameter on the comprehensive optimization target function of the surface roughness and the carbon emission is analyzed and studied by adopting a single-factor analysis method.

Influence of the feed parameters on the respective objective functions: the cutting speed is 225m/min and the back cutting amount is 0.10mm, and the feeding amount range is [0.02,0.15] mm/r. As a result, as shown in fig. 13, as the feed amount increases, the surface roughness gradually increases, the carbon emission amount gradually decreases, and the overall optimization target first decreases and then increases. Wherein the comprehensive optimization objective reaches the minimum value when the feeding amount is 0.08 mm/r.

The influence of cutting speed parameters on each objective function is that the feeding amount is set to be 0.08mm/r and the back cutting amount is set to be 0.10mm, and the cutting speed range is 80, 250 m/min. As a result, as shown in fig. 14, as the cutting speed increases, the surface roughness gradually decreases, and the carbon emission amount, which reaches the minimum value at the cutting speed of 175m/min, and the integrated optimization index, which reaches the minimum value at the cutting speed of 225m/min, both decrease and increase.

Influence of back draft parameters on the respective objective functions: the cutting speed is 225m/min and the feeding amount is 0.08mm/r, and the back cutting tool measuring range is [0.02,0.20] mm. As a result, as shown in fig. 15, the carbon emission amount gradually increased with an increase in the back draft, and the surface roughness and the comprehensive optimization index both decreased and increased, wherein the surface roughness reached the minimum at a back draft of 0.065mm, and the comprehensive optimization index reached the minimum at a back draft of 0.10 mm.

Based on the single factor influence analysis, in order to determine the quantization influence of different cutting parameters on each objective function, a standard deviation of the influence of the cutting parameters on each objective function is established, and the standard deviation of the influence of the cutting parameters on each objective function is shown in the following table 8, by comparing the standard deviation of the comprehensive optimization index, the influence degree of each parameter on the comprehensive optimization index is known to be from big to small in sequence: feed, cutting speed, back draft. Therefore, the machine tool operator can make appropriate adjustments to the hard turning process based on this result, quickly selecting more reasonable cutting parameters.

TABLE 8

By researching the optimization method of the hard turning process of the bearing, when the bearing manufacturing enterprises optimize the surface roughness and the carbon emission, the method can optimize the process parameters of the comprehensive optimization index, and has important significance for improving the production benefit of the enterprises.

The above examples represent only a few embodiments of the present invention, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.

Claims

1. The multi-target parameter optimization method for the hard turning process is characterized by comprising the following steps of:

2. The method for optimizing the multi-objective parameters of the hard turning process according to claim 1, wherein the carbon emission in the hard turning process in the step 2) is divided into direct carbon emission and indirect carbon emission, and the direct carbon emission is composed of no load, load and electric energy consumption in a tool change state in the machining process of the machine tool; the indirect carbon emission is caused by material consumption, in the hard turning process, the loss of a cutter, cutting fluid, cut-off materials and raw materials is an important influencing factor of the carbon emission in the hard turning process, the allocation of the carbon emission of the cutter, the cutting fluid, the cut-off materials and the raw materials in the total use process of the preparation system is analyzed by taking the cutting time as a unit, the consumption of the raw materials is determined by process design, the treatment of the cut-off materials is carried out after the processing, and the carbon emission optimization force of the two parts is not considered;

the obtained carbon emission model in the hard turning process is as follows:

T _c indicating the replacement period of the cutting fluid; v (V) _f Representing the consumption of the cutting fluid;representing the concentration of the cutting fluid; f (f) _t Represents the carbon emission factor, m of the cutter _t Representing the mass of the tool; t is t _p Representing a preparation time; t is t _m Representing the cutting time; t is t _ct T represents the time for single tool change _t Indicating tool life considering the number of regrind times; p (P) _uo Is the lowest idle power; d represents the most significant part of the workpiece during the polishing processA large diameter; k (K) ₁ 、K ₂ Is the relative coefficient of the rotating speed of the main shaft of the machine tool; v (V) _c Representing the turning excircle speed; t (T) _c Indicating the replacement period of the cutting fluid; v (V) _f Representing the consumption of the cutting fluid; />Representing the concentration of the cutting fluid; k. m, n, t represent coefficients related to cutting force, combined with cutting power P _c Equation (14) is determined by experimental fitting,

P _c ＝P _sp -P _u (14)

3. The method for optimizing multi-objective parameters of a hard turning process according to claim 1, wherein the method for establishing the surface roughness prediction model comprises the following steps: the cutting speed, the feeding amount and the back cutting amount which are set in the experiment are used as input layers of a generalized regression neural network, and the surface roughness measured after the workpiece is processed is used as an output layer; the input layer receives input signals, the number of neurons in the mode layer is equal to the number of samples in the training set, and nonlinear transformation is carried out on data from an input space to a mode space in the mode layer; there are two types of neurons in the summation layer: the first type of neuron calculates algebraic sum of all neurons of the pattern layer; the second type of neuron computes a weighted sum of mode layer neurons; finally, dividing two kinds of neurons of the summation layer by the output layer to obtain a predicted value of the surface roughness; in the whole operation process, the parameter smoothing factor sigma is optimized by adopting a genetic algorithm, the mean square error MSE of a neural network is used as an optimization target, global search is carried out on the smoothing factor in the value range, and the smoothing factor sigma is automatically matched to the smoothing factor value which is most suitable for the model.

4. A method for optimizing multiple target parameters of a hard turning process according to any one of claims 1 to 3, wherein the parameters in the comprehensive optimization objective function in step 4) are optimized by genetic algorithm to obtain relatively optimal cutting parameter values, namely, under the constraint condition of a model, a group of optimal process combinations are searched in cutting speed, feeding amount and back cutting amount, so that the comprehensive optimization objective function converted by a carbon emission function and a roughness model reaches the minimum value, and the specific steps are as follows:

In the middle ofRepresenting the binary encoded genetics;

b, a conversion method from a genetype to a variable:

And C, calculating the fitness of an objective function: