Micro-deformation adjustment control method for nitrocarburizing treatment of insert needle cylinder

Abstract

The invention discloses a micro-deformation adjusting and controlling method for nitrocarburizing treatment of an insert needle cylinder. The micro-deformation adjusting and controlling method comprises the steps of system construction, wide threshold value testing, narrow threshold value testing, autonomous learning and production verification. According to the method, a mode of combining a production system and an auxiliary calculation system is adopted, the concentration of H2 is controlled by adopting ammonium carbamate pre-decomposition, and a hard constraint is set to be embedded into an optimization algorithm, so that quality control is changed into beforehand prevention; the method integrates an autonomous learning model, accumulates data through a wide / narrow threshold test, establishes an accurate mapping relation between process input parameters and process output parameters, and realizes precise control of heat treatment micro-deformation through an optimization algorithm; according to the method, traditional experience trial and error are changed into data-driven optimization, so that the test times are remarkably reduced; according to the method, an adjustable weight coefficient is introduced into a loss function, so that the method has customized production capacity; the method continuously improves the production precision and the application range through the model feedback iteration data.

Description

Technical Field

[0001] This invention relates to the field of technology, and in particular to a method for adjusting and controlling micro-deformation in a needle-insertion nitrocarburizing process. Background Technology

[0002] Needle cylinders, as key components in textile machinery and precision injection molding equipment, require extremely high dimensional accuracy (micrometer level) and surface wear resistance for their inner bores. Gas nitrocarburizing treatment is a common process to improve their surface hardness, wear resistance, and fatigue resistance. However, in actual large-scale production, this process has long faced two major technical bottlenecks:

[0003] 1) The risk of hydrogen embrittlement is difficult to eliminate: Traditional processes directly introduce ammonia (NH3) and ethanol (CH3CH2OH) as nitrogen and carbon sources. Ammonia decomposes in the furnace to produce a large amount of hydrogen (H2), and the reaction is as follows: High hydrogen partial pressure causes a large number of active hydrogen atoms [H] to penetrate into the workpiece, accumulating at grain boundaries and defects, leading to a sharp decrease in material plasticity and hydrogen embrittlement. Hydrogen-embrittled parts are highly susceptible to early brittle fracture during subsequent grinding or service loading, causing batch quality accidents. Although increasing nitrogen flushing can reduce the surface hydrogen concentration, the effect is limited and cannot completely eliminate the hydrogen that has penetrated into the interior; it is only a temporary solution.

[0004] 2) Poor precision in micro-deformation control: The insert cylinder is a thin-walled cylindrical part with a complex shape. During the nitriding and carburizing heat treatment process, the coupled effects of thermal stress, structural stress, and hydrogen-induced stress can easily lead to out-of-tolerance deformation (such as outer diameter runout and increased inner hole ellipticity). The setting of traditional process parameters (such as temperature, flow rate, and drip rate) heavily relies on operator experience and lacks precise data support and feedback optimization. To find a set of parameters that simultaneously meets the requirements for hardness, layer depth, and deformation, a large number (hundreds of times) of trial and error experiments are often required, which is inefficient and costly. Even if a set of usable parameters is found, it is difficult to adaptively adjust them according to subtle changes in product requirements, resulting in poor quality stability. Workpieces with out-of-tolerance deformation often need to have their dimensions corrected by grinding the outer diameter, but this will grind away the valuable nitriding and carburizing layer, reducing the final wear resistance and service life of the workpiece.

[0005] To address the hydrogen embrittlement problem, existing technologies propose introducing carbon dioxide (CO2) to consume hydrogen within the furnace. The mechanism is based on the reverse water-gas reaction: This reaction can effectively reduce the hydrogen partial pressure, and the generated CO is an effective atmosphere for carburizing. However, directly adding CO2 will simultaneously inhibit the decomposition of ammonia, reduce the nitrogen potential, and thus weaken the nitriding effect, resulting in insufficient surface hardness.

[0006] Based on in-depth mechanistic analysis, this invention innovatively proposes using ammonium carbamate (NH2COONH4) as a pre-atmosphere source. It can completely decompose into equal proportions of NH3 and CO2 at around 150℃ (NH2COONH4→2NH3+CO2). This method cleverly provides both nitrogen and carbon sources simultaneously, and the introduction of CO2 is mild and controllable, consuming H2 without excessively inhibiting the nitriding process. Subsequently, the temperature is raised to the co-diffusion temperature, and ammonium chloride (NH4Cl) is introduced as a catalytic agent, ultimately forming a highly efficient, low-hydrogen co-diffusion atmosphere.

[0007] However, this new process is extremely complex. Numerous process input parameters, such as the ratio of ammonium carbamate to ammonium chloride, the partial pressure of the precursor gas, the temperature during co-diffusion, the NH3 flow rate, and the ethanol drop volume, collectively determine the final process output parameters, including surface hardness, layer depth, deformation, and hydrogen concentration. These parameters exhibit highly nonlinear and coupled relationships, making it virtually impossible to determine the optimal solution based on human experience alone.

[0008] Therefore, there is an urgent need for a new method that can systematically solve all the above problems. This method can not only innovatively eliminate hydrogen embrittlement from the perspective of process mechanism, but also use intelligent control methods to efficiently and accurately predict and optimize process parameters, achieve precise control of micro-deformation after nitrocarburizing treatment of insert cylinders, and ultimately achieve stable, efficient and high-quality production. Summary of the Invention

[0009] The purpose of this invention is to address the shortcomings of existing technologies by proposing a micro-deformation adjustment and control method for nitrocarburizing treatment with inserts.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] A method for adjusting and controlling micro-deformation in a needle-insertion nitrocarburizing treatment includes the following steps:

[0012] S1, System Construction

[0013] Construct a production system and an auxiliary computing system for inserting needle cylinder workpieces; the auxiliary computing system includes a database and a computing module; from the perspective of the equipment manufacturer, the production system includes an equipment setting unit, a process control unit, and a data analysis unit.

[0014] S2, Wide Threshold Test

[0015] From the perspective of process flow, the production system is divided into a test module and a production module. The test module is the data exploration and testing process used before production, including wide threshold test and narrow threshold test.

[0016] Wide threshold testing prioritizes the selection of process input parameter sets X that do not exhibit hydrogen embrittlement. n and process output parameter Yn And correct the initial historical data to obtain the corrected historical data;

[0017] S3, Narrow Threshold Test

[0018] Based on S2, the narrow threshold test uses standard process output parameters required for common workpieces. Select the corresponding process input parameter set X n After the experiment, the XY process control curve was obtained. The parameter setting, calculation and curve plotting were completed by the data analysis unit. The obtained data was further used to correct the historical data and obtain updated historical data.

[0019] S4, Self-directed learning

[0020] The computation module is a self-learning model with an input vector X and an output vector Y. The input vector X corresponds to the set of process input parameters X. n The output vector Y corresponds to the set of process input parameters X. n Actual process output parameters Y obtained from production n The mapping relationship is as follows:

[0021] Y = f(X) + ε;

[0022] and its corresponding standard process output parameters By comparing the results, the loss function is calculated:

[0023] ;

[0024] For mean squared error, the squared term ensures that the error is always positive and imposes a greater penalty on larger errors, forcing the model to prioritize solving large prediction biases.

[0025] w j As a weighting factor, it allows you to artificially emphasize or weaken certain quality indicators according to production needs;

[0026] For neural networks, θ represents all weights and biases. Training a model is essentially about finding an optimal set of θ values ​​that minimizes the measurement value of this "ruler".

[0027] S5, Production Verification

[0028] The production module manufactures workpieces based on data optimized by the calculation module.

[0029] Input the process output parameters of the standard workpiece, and based on the data optimized by the calculation module, input the corresponding process input parameters to carry out production, and monitor the process input parameters and process output parameters to obtain the process output parameters of the corresponding actual workpiece; then import the process output parameters of the actual workpiece and the process output parameters of the standard workpiece into the calculation module for verification to obtain more accurate process input parameters.

[0030] After co-infiltration treatment, the parts that have finished heat preservation are cooled with the furnace. During the cooling process, the equipment cooling fan is turned on and nitrogen is introduced at the same time to maintain the furnace pressure at 30-50 mm water column until the furnace temperature drops below 150 degrees; the furnace cover is opened and the product is lifted out.

[0031] Preferably, the device setting unit includes:

[0032] Ultrasonic cleaning machine: used for cleaning machined parts;

[0033] Dryer: Used for drying parts;

[0034] Nitriding furnace: The core equipment for nitriding and carbonizing of needle-inserted workpieces. It is equipped with a suspension mechanism for placing needle-inserted workpieces, a special decomposition plate containing auxiliary nitriding agent, a gas circulation pipeline for ammonia (NH3), a pusher feeder for dripping ethanol, a gas circulation pipeline for introducing nitrogen (N2), and a furnace cover cooling fan (used to blow nitrogen to further reduce active hydrogen [H] on the workpiece surface).

[0035] Furthermore, the auxiliary nitriding agent is made by mixing solid ammonium carbamate (NH2COONH4) and ammonium chloride (NH4Cl, nitriding catalyst) at a weight ratio of 100:1-3 and then placing it in a special decomposition tray inside the furnace.

[0036] Preferably, the process control unit includes a process for reading, measuring, acquiring, and recording process input parameters, process output parameters, and process input parameters. Process input parameters include pre-gas parameters, fluid control data, co-osmosis temperature, and pressure. Pre-gas parameters refer to the partial pressures of NH3 and CO2 pre-decomposed using a dedicated decomposition plate before co-osmosis, as well as trace amounts of HCl concentration. Fluid control data includes the NH3 flow rate and ethanol dripping rate during co-osmosis. The mechanism is as follows:

[0037] Traditional carbonitriding uses ethanol and ammonia to provide active [C] and [N] atoms during the carbonitriding process. The mechanism is a complex multi-step chemical reaction process in which CO plays a crucial role, but the ultimate goal of the entire process is to produce active nitrogen [N] and active carbon [C] atoms.

[0038] The detailed mechanism is as follows:

[0039] ① Decomposition of ammonia (NH3) (provides a nitrogen source)

[0040] Ammonia decomposes upon heating, providing reactive nitrogen atoms [N]:

[0041] ;

[0042] This reaction is reversible. The partial pressure of hydrogen (H2) in the furnace inhibits the decomposition of ammonia, so controlling the ammonia decomposition rate is actually controlling the nitrogen potential in the furnace, i.e., the activity of [N].

[0043] ②The role of ethanol (CH3CH2OH) (providing a carbon source)

[0044] The reaction pathway of ethanol is more complex, mainly involving the following two ways to generate CO, thereby providing a carbon source:

[0045] Ethanol undergoes initial cracking under the high temperature and atmosphere inside the furnace, and may react with water vapor (from the air or generated in the reaction):

[0046]

[0047] ③The core role of CO (to produce active carbon atoms [C])

[0048] The generated CO decomposes on the workpiece surface through the following reaction, providing the required active carbon atoms [C]:

[0049] ;

[0050] This reaction is the famous Boudouard reaction, which is the core and most important reaction that provides the carbon source in most gas carburizing and nitriding processes.

[0051] ④ Synergistic effect and HCN formation (a unique key to nitrocarbon co-osmosis)

[0052] In traditional single carburizing or nitriding, the above process is generally sufficient. However, the beauty of nitrocarburizing lies in the synergistic effect of nitrogen and carbon sources. Ammonia and carbon monoxide / carbon dioxide undergo a gas-phase reaction to produce hydrogen cyanide (HCN):

[0053] ;

[0054] ;

[0055] This process greatly improves the efficiency and effectiveness of co-diffusion because HCN molecules act like an "express delivery" system, simultaneously transporting [C] and [N] to the metal surface, allowing nitrogen and carbon to diffuse in synchronously and uniformly.

[0056] The above processes still cannot completely avoid hydrogen embrittlement and high-pressure cracking in the co-diffusion layer. However, this invention, through mechanistic analysis of the above processes, concludes that appropriately increasing the partial pressure of CO2 helps to reduce the H2 concentration during the co-diffusion process, thereby reducing the risk of hydrogen embrittlement.

[0057] CO2 is a weak oxidizing agent, and it can react with H2 in the following reaction (reverse reaction of water and gas):

[0058] ;

[0059] Mechanism of action:

[0060] Hydrogen consumption: CO2 reacts directly with H2 to produce H2O, which lowers the hydrogen partial pressure. Furthermore, H2O is more conducive to the degradation of ethanol.

[0061] ;

[0062] The CO produced by this reaction is a precursor to the Boudouard reaction and can also provide active carbon atoms [C]. It is much milder than directly oxidizing H2 with O2, greatly improving safety and making it less likely to oxidize the workpiece. It consumes H2 and provides a carburizing atmosphere. Moreover, this reaction is endothermic, which helps to balance the furnace temperature and reduce overheating.

[0063] However, directly adding CO2 would lead to a decrease in [N]. Therefore, it is proposed to use a pre-atmosphere condition where ammonium carbamate completely dissociates into ammonia and carbon dioxide at 150°C upon heating.

[0064] ;

[0065] Subsequently, carbonitriding was performed at around 480°C, and ammonium chloride was introduced to generate HCl to promote penetration. However, before applying the new process, it was impossible to determine the amount of ammonium carbamate to use, nor the various process input parameters required to achieve the performance data of the standard workpiece. This is why the present invention sets up a production system and an auxiliary calculation system, using a self-learning model to assist in the parameter determination process, thereby reducing the need for large-scale testing.

[0066] Preferably, the wide threshold test includes the following process:

[0067] 1) Parameter settings:

[0068] From the initial historical data in the database, filter a set X of process input parameters with varying differences within a wide range. nFor example, the partial pressure of NH3 in the pre-gas is 10-100 kPa, with a difference of 10 kPa between each group of tests; the partial pressure of CO2 is 10-100 kPa, with a difference of 10 kPa between each group of tests; and the sum of the partial pressures of NH3 and CO2 is 100 kPa; the flow rate of NH3 is 10-100 L / min, with a difference of 10 L / min between each group of tests; and the amount of ethanol added is 55-100 drops, with a difference of 5 drops / min between each group of tests.

[0069] It also detects a specific process input parameter: the H2 concentration in the nitriding furnace during production, and obtains the process output parameter Y after production. n And its corresponding process output parameters: whether hydrogen embrittlement exists, and screening out the set of process input parameters X that do not exhibit hydrogen embrittlement. n and process output parameter Y n The preliminary results showed that the safe concentration of hydrogen was 0.5%, while the hydrogen concentration in the experimental group without the addition of ammonium carbamate was consistently ≥0.8%, indicating that the improved process has significant progress.

[0070] 2) Process control curves

[0071] Filter the set of input parameters X that match the process input parameters. n and process output parameter Y n Plot the XY curve, which is the process control curve. Import the process control curve and its corresponding process input parameters, process input parameters, process output parameters and process output parameters into the database to obtain the corrected historical data.

[0072] Furthermore, the presence of hydrogen embrittlement is determined by a slow strain rate test (SSRT), i.e., by a 10... -6 -10 -7 s -1 The specimen was stretched at an extremely low strain rate, and the presence of cracking morphology or changes in ductility of the carbonitriding layer on the surface of the needle-inserted workpiece was observed. If hydrogen embrittlement was present, the carbonitriding layer would crack immediately (the workpiece body would have some ductility) (refer to ASTM F1624 standard).

[0073] Preferably, the narrow threshold test includes the following procedures:

[0074] Based on the historical correction data in the database, set the standard process output parameters required for common workpieces. For example, the surface hardness is 580-650 HV1, the co-permeation layer depth is 0.05-0.15 mm, the outer diameter runout is ≤0.02 mm, and the inner hole ellipticity is ≤0.015 mm; and based on the process control curve, the corresponding set of process input parameters X is obtained. n The production yields the corresponding actual process output parameter Y. nPlot the XY curve, which is the process control curve; verify the process control curve and the process control curve against each other, and obtain a more accurate XY curve after fitting. Import the curve and its corresponding process input parameters, process input parameters, process output parameters and process output parameters into the database to obtain updated historical data.

[0075] Preferably, the production verification of S5 further includes the following steps:

[0076] The process output parameters of the actual workpiece and the standard workpiece are then imported into the calculation module to calculate the loss function. When L(θ)≤1%, data is collected and imported into the database. When L(θ)>1%, process output parameters that do not meet the requirements are filtered out, and the gap function is calculated. Based on this, the non-standard process input parameters are obtained, and the constraint threshold is corrected to obtain more accurate process input parameters. These are then imported into the production module to continue production until the loss function L(θ) of the process output parameters of the actual workpiece and the standard workpiece is ≤1%.

[0077] Furthermore, the model parameter θ also needs to be optimized:

[0078] θ = argminL(θ);

[0079] The optimization process is as follows:

[0080] Difference function: ;

[0081] f(X) represents the prediction result using the input parameter combination X. The ideal target representing the combination of output parameters. These are weighting coefficients, allowing operators to adjust the optimization direction based on the specific requirements of this production order;

[0082] The gap function Cost(X) measures how much the predicted result f(X) differs from our ideal target Ytarget if we use the parameter combination X. The larger the gap, the higher the cost. The lowest Cost(X) represents the optimal input parameter combination X.

[0083] For example: Customer A's order requires "hardness must meet the standard, slight deformation is acceptable", then λ is set. hardness High, λ deformation Low. Customer B's order requires "deformation must be less than 0.01mm, and the lower limit of hardness must be met," so λ is set accordingly. deformation High, λ hardness If the error is low, the system will recommend a completely different process solution, which demonstrates the system's customization capabilities.

[0084] Constraints: Xlower ≤X≤X upper g(X)≤0;

[0085] X lower ≤X≤X upper Boundary constraints ensure that the recommended parameters do not exceed equipment limits (such as the maximum value of the flow pump, the temperature limit of the thermocouple) or process common sense (such as the ammonia flow rate cannot be negative).

[0086] Technical explanation: This defines the feasible domain of process parameters, ensuring that the recommended solution is not only mathematically optimal, but also physically and engineeringally feasible and safe.

[0087] ;

[0088] g(X)≤0: Process constraint;

[0089] The concentration of hydrogen is the process input parameter, which corresponds to the set of process input parameters X. This means that when the optimization algorithm searches, it will absolutely avoid recommending any solution that may cause the peak hydrogen concentration to exceed 0.5%. It directly embeds safety and reliability into the optimization algorithm, realizing proactive prevention rather than post-event alarm.

[0090] Compared with the prior art, the beneficial effects of the present invention are:

[0091] 1. This invention utilizes ammonium carbamate pre-decomposition technology to generate NH3 and CO2 in equal proportions in situ within the furnace. The CO2 is then reacted with the water gas. By continuously consuming hydrogen in the furnace, the peak H2 concentration is stably controlled below the safe threshold of 0.5%. Based on this, a wide threshold test is conducted and introduced as a constraint condition for the co-diffusion process, which accelerates the exploration of test parameters. This eliminates the risk of hydrogen embrittlement more quickly and fundamentally, significantly improves product reliability, reduces the risk of batch scrapping due to hydrogen embrittlement to zero, and greatly enhances the service safety and reliability of the product.

[0092] By directly embedding the absolute safety requirement of H2 concentration <0.5% as a hard constraint g(X)≤0 into the optimization algorithm, it means that when the model searches for the optimal parameters, it will actively avoid all areas that may cause hydrogen embrittlement risk, ensuring safety from the source; and transforming quality control from "post-event inspection" to "pre-event prevention", realizing the intelligentization and self-optimization of the production process.

[0093] 2. This invention integrates a self-learning model, accumulates data through wide / narrow threshold experiments, and establishes a precise mapping relationship between process input parameters (X) and output parameters (Y). Through optimization algorithms, the model can find the optimal parameter combination that minimizes deformation while meeting hardness and layer depth requirements, thus achieving precise control over micro-deformation during heat treatment.

[0094] 3. This invention adopts a combination of production system and auxiliary computing system, changing the traditional "experience-based trial and error" to "data-driven optimization", which significantly reduces the number of experiments. Small-batch experiments can accurately locate the optimal process window, reducing the amount of experiments by about 70%, greatly saving time, material and energy costs, and improving production efficiency.

[0095] 4. The loss function of this invention incorporates an adjustable weighting coefficient (w). j Operators can flexibly adjust the optimization direction and weighting according to the specific requirements of customer orders (such as "hardness priority" or "deformation priority"), enabling customized production and adapting to the flexible production needs of multiple varieties and small batches. For example, to prioritize ensuring deformation, λ is set. runout =2.5, λ ovality =2.0, the model will recommend a solution with better deformation and hardness slightly lower than the target value but still within the tolerance, realizing true personalized customization production.

[0096] 5. The results of each production run are fed back to the model for calculation and learning, enabling the model parameters θ to be continuously optimized and the database to be continuously expanded, thus forming a closed loop in the production verification process. This makes the production system like an experienced expert, becoming more intelligent and accurate with each use, ultimately achieving a stable product qualification rate of over 99.5%.

[0097] 6. In summary, this invention, through the deep integration of "process mechanism innovation" and "intelligent control empowerment," systematically solves the industry problem in the nitrocarbon co-infiltration treatment of needle inserts, and is an advanced technology with both significant theoretical significance and extremely high practical value. Attached Figure Description

[0098] Figure 1 This is a system architecture diagram of a micro-deformation adjustment and control method for a needle-insertion nitrocarburizing treatment proposed in this invention. Detailed Implementation

[0099] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with existing known technologies. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0100] Example 1:

[0101] A method for adjusting and controlling micro-deformation in a needle-insertion nitrocarburizing treatment includes the following steps:

[0102] S1, System Construction

[0103] Construct a production system and an auxiliary computing system for inserting needle cylinder workpieces; the auxiliary computing system includes a database and a computing module; from the perspective of the equipment manufacturer, the production system includes an equipment setting unit, a process control unit, and a data analysis unit.

[0104] The device setup unit includes:

[0105] Ultrasonic cleaning machine: used for cleaning machined parts;

[0106] Dryer: Used for drying parts;

[0107] Nitriding furnace: The core equipment for nitriding and carbonizing of needle-inserted workpieces. It is equipped with a suspension mechanism for placing needle-inserted workpieces, a special decomposition plate containing auxiliary nitriding agent, a gas circulation pipeline for ammonia (NH3), a pusher feeder for dripping ethanol, a gas circulation pipeline for introducing nitrogen (N2), and a furnace cover cooling fan (used to blow nitrogen to further reduce active hydrogen [H] on the workpiece surface).

[0108] The auxiliary nitriding agent is made by mixing solid ammonium carbamate (NH2COONH4) and ammonium chloride (NH4Cl, nitriding catalyst) at a weight ratio of 100:1-3 and then placing the mixture in a special decomposition tray inside the furnace.

[0109] The process control unit includes the processing of reading, measuring, acquiring, and inputting process input parameters, process output parameters, and process output parameters. Process input parameters include pre-gas parameters, fluid control data, co-infiltration temperature, and pressure. Pre-gas parameters refer to the partial pressures of NH3 and CO2, as well as trace amounts of HCl, pre-decomposed using a dedicated decomposition plate before co-infiltration. Fluid control data includes the NH3 flow rate and ethanol dripping rate during co-infiltration. The mechanism is as follows:

[0110] Traditional carbonitriding uses ethanol and ammonia to provide active [C] and [N] atoms during the carbonitriding process. The mechanism is a complex multi-step chemical reaction process in which CO plays a crucial role, but the ultimate goal of the whole process is to produce active nitrogen atoms [N] and active carbon atoms [C].

[0111] However, directly adding CO2 would lead to a decrease in [N]. Therefore, it is proposed to use a pre-atmosphere condition where ammonium carbamate completely dissociates into ammonia and carbon dioxide at 150°C upon heating.

[0112] Subsequently, carbonitriding was performed at around 480°C, and ammonium chloride was introduced to generate HCl to promote penetration. However, before applying the new process, it was impossible to determine the amount of ammonium carbamate to use, nor the various process input parameters required to achieve the performance data of the standard workpiece. This is why the present invention sets up a production system and an auxiliary calculation system, using a self-learning model to assist in the parameter determination process, thereby reducing the need for large-scale testing.

[0113] S2, Wide Threshold Test

[0114] From the perspective of process flow, the production system is divided into a test module and a production module. The test module is the data exploration and testing process used before production, including wide threshold test and narrow threshold test.

[0115] Wide threshold testing prioritizes the selection of process input parameter sets X that do not exhibit hydrogen embrittlement. n and process output parameter Y n And correct the initial historical data to obtain corrected historical data; 1) Parameter settings:

[0116] From the initial historical data in the database, filter a set X of process input parameters with varying differences within a wide range. n The partial pressure of NH3 in the pre-gas was 10-100 kPa, with a difference of 10 kPa between each group of tests; the partial pressure of CO2 was 10-100 kPa, with a difference of 10 kPa between each group of tests; and the sum of the partial pressures of NH3 and CO2 was 100 kPa; the flow rate of NH3 was 10-100 L / min, with a difference of 10 L / min between each group of tests; the ethanol dripping rate was 50-100 drops, with a difference of 5 drops / min between each group of tests; and the temperature was 450-600℃, with a difference of 10℃ between each group of tests.

[0117] It also detects a specific process input parameter: the H2 concentration in the nitriding furnace during production, and obtains the process output parameter Y after production. n And its corresponding process output parameters: whether hydrogen embrittlement exists, and screening out the set of process input parameters X that do not exhibit hydrogen embrittlement. n and process output parameter Y n The preliminary results showed that the safe concentration of hydrogen was 0.5%, while the hydrogen concentration in the experimental group without the addition of ammonium carbamate was consistently ≥0.8%, indicating that the improved process has significant progress.

[0118] 2) Process control curves

[0119] Filter the set of input parameters X that match the process input parameters. n and process output parameter Y nPlot the XY curve, which is the process control curve. Import the process control curve and its corresponding process input parameters, process input parameters, process output parameters and process output parameters into the database to obtain the corrected historical data.

[0120] The presence of hydrogen embrittlement is determined by a slow strain rate test (SSRT), which involves testing at a strain rate of 10... -6 -10 - 7 s -1 The specimen was stretched at an extremely low strain rate, and the presence of cracking morphology or changes in ductility of the carbonitriding layer on the surface of the needle-inserted workpiece was observed. If hydrogen embrittlement was present, the carbonitriding layer would crack immediately (the workpiece body would have some ductility) (refer to ASTM F1624 standard).

[0121] S3, Narrow Threshold Test

[0122] Based on S2, the narrow threshold test uses standard process output parameters required for common workpieces. Select the corresponding process input parameter set X n After the experiment, the XY process control curve was obtained. The parameter setting, calculation, and curve plotting were completed by the data analysis unit. The obtained data was further used to correct the historical data and obtain updated historical data.

[0123] Based on the historical correction data in the database, set the standard process output parameters required for common workpieces. The surface hardness is 580-650 HV1, the co-permeation layer depth is 0.05-0.15 mm, the outer diameter runout is ≤0.02 mm, and the inner hole ellipticity is ≤0.015 mm; and based on the process control curve, the corresponding set of process input parameters X is obtained. n The production yields the corresponding actual process output parameter Y. n Plot the XY curve, which is the process control curve; verify the process control curve and the process control curve against each other, and obtain a more accurate XY curve after fitting. Import the curve and its corresponding process input parameters, process input parameters, process output parameters and process output parameters into the database to obtain updated historical data.

[0124] S4, Self-directed learning

[0125] The computation module is a self-learning model with an input vector X and an output vector Y. The input vector X corresponds to the set of process input parameters X. n The output vector Y corresponds to the set of process input parameters X. n Actual process output parameters Y obtained from production n The mapping relationship is as follows:

[0126] Y = f(X) + ε;

[0127] and its corresponding standard process output parameters By comparing the results, the loss function is calculated:

[0128] ;

[0129] For mean squared error, the squared term ensures that the error is always positive and imposes a greater penalty on larger errors, forcing the model to prioritize solving large prediction biases.

[0130] w j As a weighting factor, it allows you to artificially emphasize or weaken certain quality indicators according to production needs;

[0131] For a neural network, θ represents all weights and biases.

[0132] Optimize the model parameter θ:

[0133] θ = argminL(θ);

[0134] The optimization process is as follows:

[0135] Difference function: ;

[0136] f(X) represents the prediction result using the input parameter combination X. The ideal target representing the combination of output parameters. These are weighting coefficients, allowing operators to adjust the optimization direction based on the specific requirements of this production order;

[0137] The gap function Cost(X) measures how much the predicted result f(X) differs from our ideal target Ytarget if we use the parameter combination X. The larger the gap, the higher the cost. The lowest Cost(X) represents the optimal input parameter combination X.

[0138] Constraints: X lower ≤X≤X upper g(X)≤0;

[0139] X lower ≤X≤X upper Boundary constraints ensure that the recommended parameters do not exceed equipment limits (such as the maximum value of the flow pump, the temperature limit of the thermocouple) or process common sense (such as the ammonia flow rate cannot be negative).

[0140] Technical explanation: This defines the feasible domain of process parameters, ensuring that the recommended solution is not only mathematically optimal, but also physically and engineeringally feasible and safe.

[0141] ;

[0142] g(X)≤0: Process constraint;

[0143] The concentration of hydrogen is the process input parameter, which corresponds to the set of process input parameters X. This means that when the optimization algorithm searches, it will absolutely avoid recommending any solution that may cause the peak hydrogen concentration to exceed 0.5%. It directly embeds safety and reliability into the optimization algorithm, realizing proactive prevention rather than post-event alarm.

[0144] S5, Production Verification

[0145] The production module manufactures workpieces based on data optimized by the calculation module.

[0146] The system inputs the process output parameters of the standard workpiece. Based on the data optimized by the calculation module, it inputs the corresponding process input parameters for production and monitors the process input and output parameters to obtain the process output parameters of the actual workpiece. The process output parameters of the actual workpiece and the standard workpiece are then imported into the calculation module to calculate the loss function. If L(θ) ≤ 1%, data is collected and imported into the database. If L(θ) > 1%, non-compliant process output parameters are filtered out, and the gap function is calculated. Based on this, non-compliant process input parameters are identified, and the constraint threshold is corrected to obtain more accurate process input parameters. These are then imported into the production module to continue production until the loss function L(θ) of the process output parameters of the actual workpiece and the standard workpiece is ≤ 1%.

[0147] After co-infiltration treatment, the parts that have finished heat preservation are cooled with the furnace. During the cooling process, the equipment cooling fan is turned on and nitrogen is introduced at the same time to maintain the furnace pressure at 30-50 mm water column until the furnace temperature drops below 150 degrees; the furnace cover is opened and the product is lifted out.

[0148] Example 2

[0149] 1. Production requirements:

[0150] Based on the basic process of Example 1, the standard process output parameters for producing the workpiece now need to be: after nitrocarburizing, the surface hardness reaches 580-650HV1, the layer depth is 0.05-0.15, the outer circle runout is 0.02, and the inner hole ellipse is within 0.015, so as to meet the service conditions of the needle holder.

[0151] 2. System Construction and Initialization

[0152] Production system: Nitriding furnaces equipped with dedicated decomposition plates and gas circulation pipelines, ultrasonic cleaners, dryers, etc.

[0153] Auxiliary computing system: Database initialization, loading existing initial historical data (including traditional process data and some preliminary exploration data).

[0154] Calculation module: Initializes a neural network model whose input vector X includes [NH3 partial pressure, CO2 partial pressure, NH3 flow rate, ethanol drop volume, co-infiltration temperature, co-infiltration time...], and output vector Y includes [surface hardness, layer depth, outer circle runout, inner hole ellipticity...].

[0155] Based on the wide threshold test determined in Example 1, the process window that does not produce hydrogen embrittlement was sampled for slow strain rate test (SSRT) after each parameter combination test. The highest H2 concentration in the furnace for all test groups (using ammonium carbamate pre-atmosphere) was 0.45%, which was lower than the safety threshold of 0.5%. The SSRT test showed that no hydrogen embrittlement cracking occurred in any of the samples, and the ductility was good.

[0156] Filter out all parameter combinations X that do not cause hydrogen embrittlement n and its corresponding output Y n The corrected historical data is then stored in the database.

[0157] 3. Narrow threshold test

[0158] Objective: To accurately match the performance indicators required by the customer within the safety window.

[0159] operate:

[0160] In correcting historical data, the target value Y requested by the customer is used. target Using a hardness of 600HV1 as a benchmark, select several sets of parameters with similar output results.

[0161] Select one set of parameters that best predicts the outcome: X n =[NH3 partial pressure: 45kPa, CO2 partial pressure: 55kPa, NH3 flow rate: 35L / min, ethanol drop rate: 75 drops / min, temperature: 480℃, time: 4h] The experiment was conducted.

[0162] result:

[0163] The actual measured result of the produced workpiece is: Y n =[Hardness: 610HV1, Layer depth: 0.11mm, Outer circle runout: 0.020mm, Ellipticity: 0.010mm].

[0164] Plot the set of XY data points, fit them together with historical data to generate a more accurate process control curve, and obtain updated historical data.

[0165] 4. Self-directed learning and optimization

[0166] Operation: The calculation module reads and updates historical data for training, optimizing the model parameters θ.

[0167] Because the customer has extremely high requirements for the amount of deformation (outer diameter runout), the operator sets a weighting coefficient λ in the loss function. hardness =1.0, λ depth =1.0, λ runout =2.5, λ ovality =2.0. This means that the model will prioritize optimizing deformation metrics, even at the cost of slightly sacrificing stiffness.

[0168] The model runs an optimization algorithm to calculate the recommended optimal parameter X for this order, under the constraint that H2 concentration < 0.5%. optimized [NH3 partial pressure: 48 kPa, CO2 partial pressure: 52 kPa, NH3 flow rate: 32 L / min, ethanol drop rate: 72 drops / min, temperature: 478℃, time: 3.8 h].

[0169] 5. Production Validation

[0170] Operation: Use recommended parameter X optimized Conduct small-batch trial production.

[0171] Real-time monitoring of process parameters confirmed that the peak H2 concentration remained stable at 0.43%.

[0172] After production is completed, workpieces are sampled for testing to obtain the actual output Y. actual [Hardness: 595HV1, Layer depth: 0.105mm, Outer circle runout: 0.016mm, Ellipticity: 0.009mm].

[0173] Feedback and iteration:

[0174] Y actual With customer standard Y target Substituting into the loss function calculation, L(θ)≈0.8% (the main difference is in hardness, but it is still within the tolerance range; the deformation is much better).

[0175] Since L(θ) ≤ 1%, production is considered successful. This high-quality (X) group... optimized Y actual The data pairs are entered into the database to further enhance the model.

[0176] Further analysis revealed that the ammonia decomposition rate during the co-infiltration process was 20-30%, the alcohol rate was 72 drops per minute, and the temperature was maintained for 4 hours, which ensured the co-infiltration size and performance requirements. Based on these parameters, mass production was carried out, and the product qualification rate reached 99.5%.

[0177] Comparative Example 1:

[0178] The same type of needle cartridge is treated using a traditional nitrocarburizing process:

[0179] 1. Process Description

[0180] The conventional gas nitrocarburization process is adopted, in which ammonia (NH3) and ethanol (CH3CH2OH) are directly introduced into the furnace at 500°C without using ammonium carbamate pre-atmosphere and without actively introducing CO2. Hydrogen embrittlement is difficult to prevent, and even increasing the flow rate of nitrogen gas is difficult to change the situation. The hydrogen concentration is >0.65%. The rest is the same as the process in Example 2.

[0181] 2. Process Design

[0182] This experiment lacks a broad experimental process and can only be conducted through randomized controlled trials. The number of trials (approximately 150 groups) is 3.5 times greater than that of Example 2 (43 groups). The optimal trial is X. n =[NH3 partial pressure: 50kPa, CO2 partial pressure: 50kPa, NH3 flow rate: 30L / min, ethanol drop rate: 80 drops / min, temperature: 490℃, time: 4h] was used for the experiment; the actual output workpiece measurement result is: Y n =[Hardness: 600HV1, Layer Depth: 0.16mm, Outer Diameter Runout: 0.024mm, Ellipticity: 0.023mm]. The issue may stem from a lack of constraints regarding hydrogen embrittlement concentration. Process parameters show a peak H2 concentration in the furnace reaching 2.1%, far exceeding the safety threshold. This can easily lead to rib breakage during outer diameter grinding, resulting in scrap. Furthermore, grinding away the nitrided layer on the outer diameter reduces the wear resistance of the insert cylinder, thus affecting its service life. This indicates that the new process in Example 2 can solve the problems of large inner hole ellipticity and large outer diameter runout during the carbonitriding process of insert cylinders, eliminating the problems caused by grinding the outer diameter after carbonitriding. This batch of parts has serious quality risks and is highly prone to breakage under stress during subsequent use, leading to the scrapping of the entire batch.

[0183] Comparative Example 2:

[0184] The calculation module was removed, and the rest of the process was the same as in Example 2. The experimental process could only be designed based on a large number of experiments, with nearly 200 sets of experiments conducted, and no optimal process was found. This is why the present invention designs a new process as in Example 2.

[0185] One set of experiments was as follows:

[0186] NH3 flow rate: 50L / min (set based on experience);

[0187] Ethanol drip rate: 90 drops / min (set based on experience);

[0188] Temperature: 500℃;

[0189] Time: 4 hours;

[0190] Product performance is as follows:

[0191] Hardness: 625HV1 (meets standards);

[0192] Layer depth: 0.12mm (meets standards);

[0193] Outer diameter runout: 0.028mm (exceeds the standard);

[0194] Inner hole ovality: 0.022mm (exceeds standard);

[0195] The peak H2 concentration inside the furnace was <0.5%, and there was no hydrogen embrittlement.

[0196] Another set of experiments was:

[0197] The furnace temperature is raised to 540 degrees in stages and held for 1.5 hours. The ammonia decomposition rate is 40-50%, and the alcohol drops 60-90 drops per minute. This can meet the requirements for hardness and layer depth, but the inner hole deformation is 0.08-0.1mm and the outer circle runout is 0.03-0.13mm.

[0198] The data from Example 2 and Comparative Examples 1-2 are shown in Table 1 below:

[0199]

[0200] The method of Example 2 is fully demonstrated to be a significant advancement and superiority in eliminating hydrogen embrittlement, controlling micro-deformation, and realizing intelligent precision manufacturing. It requires less experimentation, can quickly find optimized conditions, and can accurately locate the best process parameters.

[0201] 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 equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for adjusting and controlling micro-deformation in a needle-insertion cylinder nitrocarburizing treatment, characterized in that, Includes the following steps: S1, System Construction Construct a production system and auxiliary computing system for inserting needle cylinder workpieces; S2, Wide Threshold Test Prioritize selecting the set of process input parameters X that do not exhibit hydrogen embrittlement. n and process output parameter Y n And correct the initial historical data to obtain the corrected historical data; S3, Narrow Threshold Test Based on the standard process output parameters required for the workpiece Select the corresponding process input parameter set X n After the experiment, the XY process control curve was obtained, and the obtained data was further used to correct the historical data to obtain updated historical data. S4, Self-directed learning The mapping relationship between input vector X and output vector Y is as follows: Y = f(X) + ε; Compared with standard process output parameters By comparing the results, the loss function is calculated: ； Mean squared error; w j These are the weighting coefficients; θ represents the model parameters for all weights and biases in the neural network.

2. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 1, characterized in that, The production system includes an equipment setup unit, a process control unit, and a data analysis unit; The device setting unit includes: Ultrasonic cleaner; Dryer; Nitriding furnace: It is equipped with a suspension mechanism for placing needle-shaped workpieces, a special decomposition plate containing auxiliary nitriding agent, a gas circulation pipeline for ammonia, a propeller-type feeder for adding ethanol, a gas circulation pipeline for nitrogen, and a furnace cover cooling fan.

3. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 2, characterized in that, The auxiliary nitriding agent is made by mixing solid ammonium carbamate and ammonium chloride at a weight ratio of 100:1-3 and then placing the mixture in a special decomposition tray inside the furnace.

4. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 2, characterized in that, The process control unit includes the process of reading, measuring, acquiring and inputting process input parameters, process output parameters, process input parameters and process output parameters. The process input parameters include data such as pre-gas parameters, fluid control data, co-diffusion temperature and pressure. Pre-gas parameters include the partial pressure of NH3 and CO2, as well as the concentration of trace amounts of HCl, which are pre-decomposed using a dedicated decomposition plate before co-infiltration. Fluid control data includes the NH3 flow rate and the amount of ethanol added during co-infiltration.

5. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 1, characterized in that, The production system is divided into a testing module and a production module. The testing module is a data exploration and testing process used before production, including wide threshold testing and narrow threshold testing. The production module produces workpieces based on the data optimized by the calculation module. The wide threshold test includes the following procedures: 1) Parameter settings: The auxiliary computing system includes a database and a computing module. The computing module is a self-learning model with an input vector X and an output vector Y. The input vector X corresponds to the set of process input parameters X. n The output vector Y corresponds to the set of process input parameters X. n Actual process output parameters Y obtained from production n ; From the initial historical data in the database, filter a set X of process input parameters with varying differences within a wide range. n It also detects a specific process input parameter: the H2 concentration in the nitriding furnace during production, and obtains the process output parameter Y after production. n And its corresponding process output parameters: whether hydrogen embrittlement exists, and screening out the set of process input parameters X that do not exhibit hydrogen embrittlement. n and process output parameter Y n ; 2) Process control curves Filter the set of input parameters X that match the process input parameters. n and process output parameter Y n Plot the XY curve, which is the process control curve. Import the process control curve and its corresponding process input parameters, process input parameters, process output parameters and process output parameters into the database to obtain the corrected historical data.

6. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 5, characterized in that, The presence of hydrogen embrittlement is determined by slow strain rate testing, specifically by using a 10-strain rate test. -6 -10 -7 s -1 The specimen was stretched at an extremely low strain rate to observe whether the carbonitriding layer on the surface of the needle-inserted workpiece had crack morphology or ductility changes. If hydrogen embrittlement was present, the carbonitriding layer would crack immediately.

7. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 1, characterized in that, The narrow threshold test includes the following procedures: Based on the correction history data in the database, set the standard process output parameters required for the workpiece. And based on the process control curve, the corresponding set of process input parameters X is obtained. n The production yields the corresponding actual process output parameters Y. n The XY curve is plotted, which is the process control curve. Parameter setting, calculation and curve plotting are completed by the data analysis unit. The process control curve and the process control curve are mutually verified and fitted to obtain a more accurate XY curve. The curve and its corresponding process input parameters, process input parameters, process output parameters and process output parameters are imported into the database to obtain updated historical data.

8. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 1, characterized in that, It also includes the following steps: The process output parameters of the actual workpiece and the standard workpiece are then imported into the calculation module to calculate the loss function. When L(θ)≤1%, data is collected and imported into the database. When L(θ)>1%, process output parameters that do not meet the requirements are filtered out, and the gap function is calculated. Based on this, the non-standard process input parameters are obtained, and the constraint threshold is corrected to obtain more accurate process input parameters. These are then imported into the production module to continue production until the loss function L(θ) of the process output parameters of the actual workpiece and the standard workpiece is ≤1%.

9. The micro-deformation adjustment and control method for nitrocarburizing treatment of needle-insertion cylinders according to claim 8, characterized in that, The model parameter θ also needs to be optimized: θ = argminL(θ); The optimization process is as follows: Difference function: ; f(X) represents the prediction result using the input parameter combination X. The ideal target representing the combination of output parameters. These are the weighting coefficients; Constraints: Boundary constraints: X lower ≤X≤X upper ; Process constraint: g(X)≤0; in , The process input parameter is the concentration of hydrogen gas. 0.5 represents a safe hydrogen concentration of 0.5%.

Images