CN104589606B

CN104589606B - A kind of injection molding process on-line monitoring method

Info

Publication number: CN104589606B
Application number: CN201510022875.6A
Authority: CN
Inventors: 张云; 周华民; 黄志高; 毛霆; 高煌; 李德群; 周循道
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2015-01-16
Filing date: 2015-01-16
Publication date: 2015-10-28
Anticipated expiration: 2035-01-16
Also published as: CN104589606A

Abstract

The invention discloses an online monitoring method for injection molding process, belonging to the field of process monitoring, which includes collecting process variables in the injection molding process, calculating and obtaining the feature vector of the process variables of production batches, and then selecting several qualified The monitoring template SP ^model and the monitoring control limit D ^model are calculated based on the characteristic vector of the process variable of the batch, and are obtained by calculating the monitoring template, monitoring control limit and the characteristic vector SP _z of the z-th batch in the current production process The monitoring quantity D _z and monitoring index T _z of the zth batch, when T _z ≤ 1, it is judged that the production process of the zth batch is qualified, and when T _z > 1, the production process of the zth batch is judged abnormal. The method of the present invention does not require additional investment in equipment and funds, has low economic costs, avoids the problems of batch data alignment and Gaussian assumptions in the PCA method, and can well adapt to the system drift of the injection molding process, and can also control injection molding production as expected process.

Description

A method for online monitoring of injection molding process

技术领域technical field

本发明属于过程监控领域，更具体地，涉及一种基于多时段统计模式分析的注塑成型过程在线监控方法。The invention belongs to the field of process monitoring, and more specifically relates to an online monitoring method for injection molding process based on multi-period statistical pattern analysis.

背景技术Background technique

批次生产过程，即间歇生产过程，是指在有限的时间内、以预先设计好的工序将原材料加工成符合质量要求的产品的过程。批次生产过程通常由多道工序组成，每道工序有多个控制变量。在批次生产过程中，由于每道工序的运行时间不确定性、反应复杂性、变量的时变性、变量与变量之间相互关联性，从而很难通过基于机理模型来建立起批次生产过程与产品质量之间的关系。The batch production process, that is, the batch production process, refers to the process of processing raw materials into products that meet quality requirements within a limited time and with pre-designed procedures. Batch production processes usually consist of multiple processes, each with multiple control variables. In the batch production process, due to the uncertainty of the running time of each process, the complexity of the response, the time-varying nature of variables, and the interrelationship between variables, it is difficult to establish a batch production process based on a mechanism model. relationship with product quality.

注塑成型过程是典型的批次生产过程，注塑成型过程一般包括合模阶段、注射阶段、保压阶段、塑化阶段、射退阶段、冷却阶段和开模取出阶段。在注塑成型的全过程中，需要关注的过程变量包括系统压力、螺杆行程、螺杆转速、螺杆速度、料筒温度、喷嘴温度、模具温度、塑化阶段的塑化背压、型腔压力以及喷嘴压力。注射阶段熔体的充模速度对塑料熔体分子排列和剪切应力有着直接的影响，从而影响制品质量。在保压阶段，熔体在高压下慢速流动，模腔内熔体被压缩增密，制品逐渐成型，保压压力是需要关注的。塑化质量取决于塑化阶段过程的工艺参数以及塑料原料的热物理性能和流变性能，制品质量与料筒中熔体质量有着密切的关系。射退阶段关心的是螺杆行程。模腔内的冷却过程使制品具有一定的刚度和强度，但是过长的冷却时间对制品质量作用较小。整个过程中，全程最需要监控的变量为螺杆行程和各种压力，其中压力具体是指，注射阶段的系统压力、保压阶段的系统压力、塑化阶段的塑化背压等。注塑成型制品质量与批次生产过程中的各个工序的多个变量有着直接或者间接的关系。The injection molding process is a typical batch production process. The injection molding process generally includes a mold clamping stage, an injection stage, a pressure holding stage, a plasticizing stage, an ejection stage, a cooling stage, and a mold opening and removal stage. In the whole process of injection molding, the process variables that need to be paid attention to include system pressure, screw stroke, screw speed, screw speed, barrel temperature, nozzle temperature, mold temperature, plasticizing back pressure in the plasticizing stage, cavity pressure and nozzle pressure. The filling speed of the melt in the injection stage has a direct impact on the molecular arrangement and shear stress of the plastic melt, thus affecting the quality of the product. In the holding stage, the melt flows slowly under high pressure, the melt in the mold cavity is compressed and densified, and the product is gradually formed. The holding pressure needs to be paid attention to. The quality of plasticization depends on the process parameters of the plasticization stage and the thermophysical and rheological properties of the plastic raw materials. The quality of the product is closely related to the quality of the melt in the barrel. The ejection stage is concerned with the screw stroke. The cooling process in the mold cavity makes the product have a certain rigidity and strength, but too long cooling time has little effect on the quality of the product. During the whole process, the variables that need to be monitored most throughout the whole process are the screw stroke and various pressures. The pressure specifically refers to the system pressure in the injection stage, the system pressure in the pressure holding stage, and the plasticizing back pressure in the plasticizing stage. The quality of injection molded products is directly or indirectly related to multiple variables in each process in the batch production process.

在注塑成型生产过程中，由于材料、环境温度、以及机器等因素的波动，引起注塑成型过程批次间的波动，从而导致制品质量有一定波动性。注塑成型批次生产过程的在线监控，是对于给定的注塑机、材料、模具、以及对应的工艺参数进行的监控，通过监控注塑成型生产过程的过程变量，可达到监控注塑成型过程稳定性以及预测注塑成型制品质量的目的。In the injection molding production process, due to fluctuations in materials, ambient temperature, and machines, etc., the fluctuations between batches in the injection molding process are caused, resulting in certain fluctuations in product quality. The online monitoring of the injection molding batch production process is the monitoring of the given injection molding machine, material, mold, and corresponding process parameters. By monitoring the process variables of the injection molding production process, the stability of the injection molding process can be monitored and The purpose of predicting the quality of injection molded products.

对于注塑过程监控，传统的主成分分析(Principal ComponentAnalysis,PCA)方法在注塑成型生产过程中存在一些问题：(1)其存在批次时间对齐的问题；(2)该方法对于过程变量在批次方向上高斯分布的假设在实际生产过程中并不一定成立，使该方法存在较大的不准确性；(3)随着内部系统偏移以及外部不可控因素的影响，PCA模型往往不能及时更新反映当前注塑过程的特性而出现误报警。For injection molding process monitoring, the traditional Principal Component Analysis (PCA) method has some problems in the injection molding production process: (1) it has the problem of batch time alignment; The assumption of Gaussian distribution in the direction is not necessarily true in the actual production process, which makes the method have great inaccuracy; (3) With the influence of internal system deviation and external uncontrollable factors, the PCA model often cannot be updated in time False alarms occur due to the characteristics of the current injection molding process.

发明内容Contents of the invention

针对现有技术的以上缺陷或改进需求，本发明提供了一种注塑成型过程中在线监控方法，其目的在于建立一套全新的实时监控注塑成型的模型，用于监控生产过程中不同批次的不同阶段中的不同过程变量，能反映注塑成型制品质量的波动，还能在监控到过程变量波动太大时给予反馈并进行自动更新，该模型精确度高，可以保证不同批次间产品质量的一致性，由此解决现有注塑过程监控方法存在模型精度不高、不能及时更新易出现误报的技术问题。Aiming at the above defects or improvement needs of the prior art, the present invention provides an online monitoring method in the injection molding process, the purpose of which is to establish a new real-time monitoring injection molding model for monitoring different batches of Different process variables in different stages can reflect the fluctuation of the quality of injection molding products, and can also give feedback and automatically update when the monitored process variables fluctuate too much. The model has high accuracy and can ensure the quality of products between different batches. Consistency, thereby solving the technical problems of the existing injection molding process monitoring methods, such as low model accuracy, failure to update in time and prone to false positives.

为实现上述目的，本发明提供了一种注塑成型过程在线监控方法，用于监控注塑机生产过程，其特征在于，包括如下步骤：In order to achieve the above object, the present invention provides an online monitoring method for injection molding process, which is used to monitor the production process of injection molding machine, which is characterized in that it includes the following steps:

S1：采集注塑成型过程中第i个批次的第m个子阶段的第j个过程变量x_i,m,j，其中，i＝1,2,…,I；j＝1,2,…，J；m＝1,2,…,M，且I为批次总数，J为过程变量总数，M为子阶段总数，所述子阶段至少包括注射阶段、保压阶段、塑化阶段以及射退阶段，所述过程变量至少包括系统压力和螺杆行程；S1: Collect the j-th process variable x _i,m,j of the m-th sub-stage of the i-th batch in the injection molding process, where, i=1,2,…,I; j=1,2,…, J; m=1,2,...,M, and I is the total number of batches, J is the total number of process variables, M is the total number of sub-stages, and the sub-stages include at least the injection stage, the pressure holding stage, the plasticizing stage and the ejection stage stage, the process variables include at least system pressure and screw stroke;

S2：先计算步骤S1中所述x_i,m,j的特征量和特征量向量，所述特征量包括过程变量的最小值σ_i,m,j,1、最大值σ_i,m,j,2、平均值μ_i,m,j、方差ε_i,m,j、偏度γ_i,m,j、峰度κ_i,m,j、信息熵所述特征量向量如下：S2: First calculate the feature quantity and feature quantity vector of x _i,m,j mentioned in step S1, the feature quantity includes the minimum value σ _i,m,j,1 and the maximum value σ _i,m,j of the process variable _,2 , mean value μ _i,m,j , variance ε _i,m,j , skewness γ _i,m,j , kurtosis κ _i,m,j , information entropy The feature quantity vector is as follows:

以上各个特征量以及所述特征量向量中，i＝1,2,…，I；j＝1,2，…，J；m＝1,2,…,M，且过程变量总数J≥2，子阶段总数M≥4；In each of the above feature quantities and the feature quantity vector, i=1,2,...,I; j=1,2,...,J; m=1,2,...,M, and the total number of process variables J≥2, The total number of sub-stages M≥4;

接着计算获得第i个批次的特征量向量SP_i，所述SP_i如下：Then calculate and obtain the feature quantity vector SP _i of the i-th batch, and the SP _i is as follows:

SP_i＝vec[SP_i,m,j]SP _i =vec[SP _i,m,j ]

＝[SP_i，1，1，SP_i，1，2，…，SP_i，1，J，SP_i，2，1，SP_i，2，2，…,SP_i，2，J，…，SP_i，M，1，SP_i，M，2，…，SP_i，M，J]^T =[SP _i,1,1 ,SP _i,1,2 ,...,SP _i,1,J ,SP _i,2,1 ,SP _i,2,2 ,...,SP _i,2,J ,..., SP _{i, M, 1} , SP _{i, M, 2} ,..., SP _{i, M, J} ] ^T

其中，vec[·]表示将数据排成列向量形式；Among them, vec[ ] means to arrange the data into a column vector form;

S3：计算获得B个合格批次的监控模板SP^model和B个合格批次的监控控制限D^model，分别为如下：S3: Calculate and obtain the monitoring template SP ^model of B qualified batches and the monitoring control limit D ^model of B qualified batches, which are as follows:

${SP SP}^{mod mod e e l l} = = \frac{11}{B B} {Σ Σ}_{b b = = 11}^{B B} {SP SP}_{b b}^{t t r r a a i i n no}$

D^model＝(1-β)||SP^model||D ^model ＝(1-β)||SP ^model ||

以上式中，||·||为向量2范数计算，β为监控可靠度，β在范围(0，1)内任意取值，B为小于I的批次数量，为总共B个批次中第b个批次的特征量向量；In the above formula, ||·|| is the vector 2 norm calculation, β is the monitoring reliability, β is any value in the range (0, 1), B is the number of batches less than 1, is the feature vector of the bth batch in a total of B batches;

S4：计算当前生产过程中的第z个批次的监控量D_z和监控指标T_z，分别如下：S4: Calculate the monitoring quantity D _z and monitoring index T _z of the zth batch in the current production process, respectively as follows:

D_z＝||SP_z-SP^model||D _z ＝||SP _z -SP ^model ||

${T T}_{z z} = = \frac{{D D.}_{z z}}{{D D.}^{mod mod e e l l}}$

以上两式中，SP_z为第z个批次的特征量向量，SP^model为B个合格批次的监控模板和B个合格批次监控控制限D^model；In the above two formulas, SP _z is the feature quantity vector of the zth batch, and SP ^model is the monitoring template of B qualified batches and the monitoring control limit D ^model of B qualified batches;

S5：进行在线监控判断，当T_z≤1时，判定该第z个批次生产过程合格；当T_z＞1时，判定该第z个批次生产过程异常，实现对注塑成型过程的在线监控。S5: Carry out online monitoring and judgment. When T _z ≤ 1, it is judged that the production process of the z-th batch is qualified; when T _z > 1, it is judged that the production process of the z-th batch is abnormal, and the online injection molding process is realized. monitor.

进一步的，还包括步骤S6，所述步骤S6为对B个合格批次的监控模板SP^model和B个合格批次的监控控制限D^model进行更新，以新监控模板SP^new-model替换B个合格批次的监控模板SP^model，新监控模板SP^new-model如下：Further, step S6 is also included, the step S6 is to update the monitoring template SP ^model of B qualified batches and the monitoring control limit D ^model of B qualified batches, and replace B with the new monitoring template SP ^new-model The monitoring template SP ^model of qualified batches and the new monitoring template SP ^new-model are as follows:

SP^new-model＝(1-λ)·SP^model+λ·SP_z SP ^new-model = (1-λ) SP ^model + λ SP _z

式中，SP_z为第z个批次的特征量向量，SP^model为B个合格批次的监控模板，λ为学习因子，为任意非负数；In the formula, SP _z is the feature vector of the zth batch, SP ^model is the monitoring template of B qualified batches, and λ is the learning factor, which is any non-negative number;

以新监控控制限D^new-model替换B个合格批次的监控控制限D^model，新监控控制限D^new-model如下：Replace the monitoring control limit D model of B qualified batches with the new monitoring control limit D ^new - ^model . The new monitoring control limit D ^new-model is as follows:

D^new-model＝(1-β)||SP^new-model||D ^new-model ＝(1-β)||SP ^new-model ||

式中，SP^new-model为新监控模板，β为监控可靠度。In the formula, SP ^new-model is the new monitoring template, and β is the monitoring reliability.

进一步的，所述过程变量的最小值σ_i,m,j,1、最大值σ_i,m,j,2、平均值μ_i,m,j、方差ε_i,m,j、偏度γ_i,m,j、峰度κ_i,m,j、信息熵分别采用如下公式计算：Further, the minimum value σ _i,m,j,1 , maximum value σ _i,m,j,2 , mean value μ _i,m,j , variance ε _i,m,j , skewness γ of the process variable _i,m,j , kurtosis κ _i,m,j , information entropy Calculated using the following formulas:

σ_i,m,j,1＝min[x_i,m,j(α)]σ _i,m,j,1 = min[ _xi,m,j (α)]

σ_{i，m，j，2}＝max[x_i，m，j(α)]σ _{i, m, j, 2} = max[x _{i, m, j} (α)]

${μ μ}_{i i,, m m,, j j} = = \frac{11}{{K K}_{i i,, m m,, j j}} {Σ Σ}_{α α = = 11}^{{K K}_{i i,, m m,, j j}} {x x}_{i i,, m m,, j j} ((α α))$

${ϵ ϵ}_{i i,, m m,, j j} = = \frac{11}{{K K}_{i i,, m m,, j j} - - 11} {Σ Σ}_{α α = = 11}^{{K K}_{i i,, m m,, j j}} {(({x x}_{i i,, m m,, j j} ((α α)) - - {μ μ}_{i i,, m m,, j j}))}^{22}$

${γ γ}_{i i,, m m,, j j} = = \frac{\frac{11}{{K K}_{i i,, m m,, j j}} {Σ Σ}_{α α = = 11}^{{K K}_{i i,, m m,, j j}} {(({x x}_{i i,, m m,, j j} ((α α)) - - {μ μ}_{i i,, m m,, j j}))}^{22}}{{[[\frac{11}{{K K}_{i i,, m m,, j j}} {Σ Σ}_{α α = = 11}^{{K K}_{i i,, m m,, j j}} {(({x x}_{i i,, m m,, j j} ((α α)) - - {μ μ}_{i i,, m m,, j j}))}^{22}]]}^{33 / / 22}}$

${κ κ}_{i i,, m m,, j j} = = \frac{\frac{11}{{K K}_{i i,, m m,, j j}} {Σ Σ}_{α α = = 11}^{{K K}_{i i,, m m,, j j}} {(({x x}_{i i,, m m,, j j} ((α α)) - - {μ μ}_{i i,, m m,, j j}))}^{44}}{{[[\frac{11}{{K K}_{i i,, m m,, j j}} {Σ Σ}_{α α = = 11}^{{K K}_{i i,, m m,, j j}} {(({x x}_{i i,, m m,, j j} ((α α)) - - {μ μ}_{i i,, m m,, j j}))}^{22}]]}^{22}} - - 33$

式中，α＝1,2,…,K_i,m,j，所述K_i,m,j为x_i,m,j的长度。In the formula, α=1,2,...,K _i,m,j , where K _i,m,j is the length of x _i,m,j .

进一步的，采用注塑机自带的传感器采集所述过程变量。注塑机本身自带计时器、压力传感器、温度传感器、位置传感器，其中，压力传感器具有多个，分别分布在喷嘴处，型腔壁上，系统动力传导处。温度传感器分布在料筒的各段，模具的型腔壁上，以及喷嘴处。位置传感器位于螺杆附近，用于测量螺杆在不同阶段时候的位移。Further, the process variable is collected by using a sensor attached to the injection molding machine. The injection molding machine itself has its own timer, pressure sensor, temperature sensor, and position sensor. Among them, there are multiple pressure sensors, which are respectively distributed at the nozzle, the cavity wall, and the power transmission of the system. The temperature sensors are distributed in each section of the barrel, on the cavity wall of the mold, and at the nozzle. Position sensors are located near the screw to measure the displacement of the screw at different stages.

进一步的，所述过程变量还包括螺杆转速、螺杆行进速度、料筒温度、喷嘴温度、模具温度、塑化阶段的塑化压力、型腔压力、喷嘴压力、注射阶段的充模速度以及保压阶段的保压压力。Further, the process variables also include screw speed, screw travel speed, barrel temperature, nozzle temperature, mold temperature, plasticizing pressure in the plasticizing stage, cavity pressure, nozzle pressure, mold filling speed in the injection stage and holding pressure The packing pressure of the stage.

进一步的，所述特征量还包括分位点、差值以及积分。Further, the feature quantity also includes quantile points, differences and integrals.

进一步的，所述子阶段还包括合模阶段、冷却阶段和开模阶段。各个子阶段按照时间顺序连串起来组成一个周期，一个周期即是一个批次。根据注塑成型过程内在机理，可选择监控一个或者多个子阶段的过程数据。Further, the sub-stages also include a mold clamping stage, a cooling stage and a mold opening stage. Each sub-stage is connected in series according to time to form a cycle, and a cycle is a batch. According to the intrinsic mechanism of the injection molding process, one or more sub-stage process data can be selected to be monitored.

总体而言，通过本发明所构思的以上技术方案，能够取得下列有益效果：Generally speaking, through the above technical solutions conceived by the present invention, the following beneficial effects can be obtained:

1、本发明中数据采集不需要额外增加传感器，直接采用注塑机自带的传感器即可采集所需数据，无需额外投入设备和资金，经济成本低。1. In the present invention, no additional sensors are required for data collection, and the required data can be collected directly by using the sensors that come with the injection molding machine, without additional investment in equipment and funds, and the economic cost is low.

2、本发明方法中采用计算注塑成型批次过程数据的特征量向量的表征方法来监控过程变量，可以很好的表征注塑成型批次过程的特点，还有效的避免了PCA方法中批次数据对齐问题以及高斯假设的问题，保证了本发明方法相对于PCA方法更准确。2. In the method of the present invention, the characterization method of calculating the feature quantity vector of the injection molding batch process data is used to monitor the process variables, which can well characterize the characteristics of the injection molding batch process, and effectively avoid the batch data in the PCA method. The alignment problem and the Gaussian assumption guarantee that the method of the present invention is more accurate than the PCA method.

3、本发明方法在建模阶段不需要大量的训练样本数据，只需要若干个甚至是一个正常批次的过程数据作为训练样本，监控模型的更新过程简单，直接以最新监控模板和最新监控控制限分别替换当前监控模板和当前监控控制限，即能完成模型的更新，模型的动态更新可以很好的适应注塑成型过程的系统漂移。3. The method of the present invention does not require a large amount of training sample data in the modeling stage, and only needs several or even a normal batch of process data as training samples. The update process of the monitoring model is simple, and the latest monitoring template and the latest monitoring control are directly used. The current monitoring template and the current monitoring control limit are replaced by the current monitoring control limit, that is, the update of the model can be completed, and the dynamic update of the model can well adapt to the system drift of the injection molding process.

4、本发明方法中建立的在线监控指标与注塑生产过程直接相关，能反应注塑成型中制品质量的波动，还可以通过调整监控可靠度的大小，灵活方便的控制注塑生产过程，使所得的制品质量满足预期要求。4. The online monitoring index established in the method of the present invention is directly related to the injection molding production process, which can reflect the fluctuation of product quality in injection molding, and can also flexibly and conveniently control the injection molding production process by adjusting the size of the monitoring reliability, so that the obtained product The quality meets the expected requirements.

附图说明Description of drawings

图1是本发明实施例提供的注塑成型过程在线监控方法的流程示意图；1 is a schematic flow diagram of an online monitoring method for injection molding process provided by an embodiment of the present invention;

图2是本发明实施例提供的注塑成型过程某个批次过程采样数据曲线示意图；Fig. 2 is a schematic diagram of a certain batch process sampling data curve of the injection molding process provided by the embodiment of the present invention;

图3是本发明实施例提供的注塑成型过程曲线监控示意图。Fig. 3 is a schematic diagram of the curve monitoring of the injection molding process provided by the embodiment of the present invention.

具体实施方式Detailed ways

为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。此外，下面所描述的本发明各个实施方式中所涉及到的技术特征只要彼此之间未构成冲突就可以相互组合。In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not constitute a conflict with each other.

图1是本发明实施例提供的注塑成型过程在线监控方法的流程示意图，下面结合图1对该方法进行详细说明。FIG. 1 is a schematic flowchart of an online monitoring method for injection molding process provided by an embodiment of the present invention. The method will be described in detail below with reference to FIG. 1 .

(1)采集注塑成型过程中的过程变量(1) Acquisition of process variables in the injection molding process

本实施例中，注塑机型号为恩瑞德NC60，注塑材料为聚丙烯，生产的制品为白色透明的盒形制品。数据采集中取采样频率f＝333.33Hz，即每3毫秒采样一次，利用注塑机现有的传感器对注塑成型过程进行过程变量采集，采集的过程变量个数J＝2，过程变量为系统压力和螺杆位置，注塑过程的子阶段个数M＝4，子阶段分别为注射阶段、保压阶段、塑化阶段和射退阶段。图2是注塑成型过程某一批次的实时采样数据曲线示意图，从图中可知，随着不同子阶段的变化，系统压力先增大后减小，最后持平，螺杆位置先减小后增大。表1是注塑成型过程第10个批次的数据采集表。该表中，子阶段值1、2、3和4分别表示注射阶段、保压阶段、塑化阶段和射退阶段，对应的采样个数分别为310、163、1733和114。每个子阶段采样个数不同的原因是，在同样的采样频率下每个子阶段的持续时间不同。In this embodiment, the model of the injection molding machine is Enrad NC60, the injection molding material is polypropylene, and the products produced are white transparent box-shaped products. In the data collection, the sampling frequency f=333.33Hz is taken, that is, sampling is performed every 3 milliseconds. The existing sensor of the injection molding machine is used to collect the process variables of the injection molding process. The number of collected process variables is J=2, and the process variables are system pressure and The position of the screw, the number of sub-stages of the injection molding process M = 4, and the sub-stages are injection stage, pressure holding stage, plasticizing stage and ejection stage. Figure 2 is a schematic diagram of the real-time sampling data curve of a certain batch of injection molding process. It can be seen from the figure that with the change of different sub-stages, the system pressure first increases and then decreases, and finally stays the same. The screw position first decreases and then increases . Table 1 is the data collection table of the 10th batch in the injection molding process. In this table, the sub-stage values 1, 2, 3, and 4 represent the injection stage, pressure holding stage, plasticizing stage, and ejection stage, respectively, and the corresponding sampling numbers are 310, 163, 1733, and 114, respectively. The reason for the different number of samples in each sub-stage is that the duration of each sub-stage is different under the same sampling frequency.

表1是注塑成型过程第10个批次的数据采集表Table 1 is the data collection table of the 10th batch of injection molding process

采样序号Sampling number 子阶段sub-stage 系统压力(0.1Bar)System pressure (0.1Bar) 螺杆位置(0.1毫米)Screw position (0.1 mm) 11 11 1111 844844 22 11 99 844844 ……... ……... ……... ……... 309309 11 383383 304304 310310 11 385385 302302 311311 22 386386 297297 312312 22 389389 293293 ……... ……... ……... ……... 472472 22 8787 218218 473473 22 8181 219219 474474 33 7373 220220 475475 33 7070 221221 ……... ……... ……... ……... 22052205 33 282282 801801 22062206 33 278278 801801 22072207 44 161161 804804 22082208 44 134134 804804 ……... ……... ……... ……... 23192319 44 24twenty four 842842 23202320 44 24twenty four 843843

(2)过程变量的特征量和特征量向量计算(2) Calculation of eigenvalues and eigenvalue vectors of process variables

对于第i个批次的第m个子阶段中的第j个过程变量的采样数据x_i，m，j进行特征量计算，得到由等长的多个特征量组成的特征量向量SP_i，m，j：For the sampling data x _{i, m, j} of the j-th process variable in the m-th sub-stage of the i-th batch, the feature quantity calculation is performed, and the feature quantity vector SP _{i, m} composed of multiple feature quantities of equal length is obtained _{, j} :

SP_i，m，j＝[σ_{i，m，j，1}，σ_{i，m，j，2}，μ_i，m，j，ε_i，m，j，γ_i，m，j，κ_i，m，j，∈_i，m，j]^T，SP _{i, m, j} = [σ _{i, m, j, 1} , σ _{i, m, j, 2} , μ _{i, m, j} , ε _{i, m, j} , γ _{i, m, j} , κ _{i, m, j} , ∈ _{i, m, j} ] ^T ,

式中，σ_{i，m，j，1}，σ_{i，m，j，2}，μ_i，m，j，ε_i，m，j，γ_i，m，j，κ_i，m，j，∈_i，m，j分别为x_i，m，j的最小值、最大值、平均值、方差、偏度、峰度和信息熵。In the formula, σ _{i, m, j, 1} , σ _{i, m, j, 2} , μ _{i, m, j} , ε _{i, m, j} , γ _{i, m, j} , κ _{i, m, j} , ∈ _{i, m, j} are the minimum, maximum, average, variance, skewness, kurtosis and information entropy of x _{i, m} , j respectively.

其中，in,

σ_i,m,j,1＝min[x_i,m,j(α)]σ _i,m,j,1 = min[ _xi,m,j (α)]

σ_i,m,j,2＝max[x_i,m,j(α)]σ _i,m,j,2 = max[ _xi,m,j (α)]

${ϵ ϵ}_{i i,, m m,, j j} = = \frac{11}{{K K}_{i i,, m m,, j j} - - 11} {Σ Σ}_{α α = = 11}^{{K K}_{i i,, m m,, j j}} {(({x x}_{i i,, m m,, . . j j} ((α α)) - - {μ μ}_{i i,, m m,, j j}))}^{22}$

以上各式中，α＝1,2,…,K_i,m,j，所述K_i,m,j为x_i,m,j的长度。In the above formulas, α=1,2,...,K _i,m,j , where K _i,m,j is the length of x _i,m,j .

将每一批次采样数据按照子阶段和过程变量展开，得到每一批次采样数据的特征量向量SP_i，即SP_i＝vec[SP_m，j]，其中m＝1，2，…4；j＝1，2。Expand each batch of sampling data according to sub-stages and process variables to obtain the feature vector SP _i of each batch of sampling data, that is, SP _i =vec[SP _{m, j} ], where m=1, 2,...4 ;j=1,2.

表2第10个批次的采样数据的特征量计算结果Table 2 Calculation results of feature quantities of the tenth batch of sampled data

以第10个批次的采样数据为例，其特征量计算的结果如表2所示。表2中，每一行代表一个子阶段的一个过程变量的特征量向量，如第1行代表注塑充模阶段系统压力的特征量向量,表示为SP_10，1，1＝[0.00，262.02，386.00，14495.20，-0.93，2.50，0.10]^T，最终，SP₁₀为表中所有数据按列向量展开得到的特征量向量。Taking the sampling data of the 10th batch as an example, the calculation results of its feature quantities are shown in Table 2. In Table 2, each line represents a feature vector of a process variable in a sub-stage, such as the first line represents the feature vector of the system pressure in the injection mold filling stage, expressed as SP 10, 1, 1 = [0.00, _262.02 , 386.00 , 14495.20, -0.93, 2.50, 0.10] ^T , finally, SP ₁₀ is the feature vector obtained by expanding all the data in the table according to the column vector.

(3)计算获得B个合格批次的监控模板SP^model和B个合格批次的监控控制限D^model。(3) Calculate and obtain the monitoring template SP ^model of B qualified batches and the monitoring control limit D ^model of B qualified batches.

首先，将采集数据时候获得的关于生产合格的制品的数据中选取B个批次的数据组成训练样本集合X^train，对每个训练样本b＝1，2，…，B，计算获得总共B个批次中第b个批次特征量向量获得B个由特征量向量组成的训练样本特征量向量矩阵SP^train，如下所示：First of all, select B batches of data from the data about the production of qualified products obtained when collecting data to form a training sample set X ^train , for each training sample b=1, 2,..., B, calculate and obtain the feature vector of the bth batch in a total of B batches Obtain B training sample feature vector matrix SP ^train composed of feature vectors, as follows:

${SP SP}^{t t r r a a i i n no} = = r r o o w w [[{(({SP SP}_{b b}^{t t r r a a i i n no}))}^{T T}]]$

式中，(·)^T为向量转置运算，为总共B个批次中第b个批次的特征量向量；In the formula, ( ) ^T is the vector transpose operation, is the feature vector of the bth batch in a total of B batches;

接着，计算B个批次的特征量监控模板SP^model，如下：Next, calculate the feature quantity monitoring template SP ^model of B batches, as follows:

式中，B为所选取的建模批次个数，为B个建模批次中第b个批次的特征量向量。In the formula, B is the number of modeling batches selected, is the feature vector of the b-th batch in the B modeling batches.

再计算B个合格批次的监控控制限D^model，如下：Then calculate the monitoring control limit D ^model of B qualified batches, as follows:

D^model＝(1-β)||SP^model||D ^model ＝(1-β)||SP ^model ||

式中，||·||为向量2范数计算，β为监控可靠度，其在范围(0，1)内任意取值。In the formula, ||·|| is the vector 2 norm calculation, and β is the monitoring reliability, which can take any value within the range (0, 1).

本实施例中，取B＝5个合格批次的数据建模，计算得到SP^model,取监控可靠度β＝0.99，从而计算获得D^model＝452.5085。即建模主要是获得监控模板SP^model和监控控制限D^model，以用于后续的比对判断。In this embodiment, the data of B=5 qualified batches are modeled, and SP ^model is calculated, and the monitoring reliability β=0.99 is calculated, so that D ^model =452.5085 is calculated. That is, the modeling is mainly to obtain the monitoring template SP ^model and the monitoring control limit D ^model for subsequent comparison and judgment.

(4)计算当前生产过程中的第z个批次的监控量D_z和监控指标T_z，分别如下：(4) Calculate the monitoring quantity D _z and monitoring index T _z of the zth batch in the current production process, respectively as follows:

对于被实时监测的当前生产过程中的第z个批次的全周期数据X_z，计算获得全周期数据X_z的监控量D_z，其计算过程如下：For the full-cycle data X z of the _{z-th batch in the current production process that is monitored in real time, the monitoring amount D z} _{of the full-cycle data X z} _is calculated, and the calculation process is as follows:

D_z＝||SP_z-SP^model||D _z ＝||SP _z -SP ^model ||

式中，SP_z为第z个批次的特征量向量，SP^model为监控模板。In the formula, SP _z is the feature vector of the zth batch, and SP ^model is the monitoring template.

再计算监控指标T_z，其公式如下：Then calculate the monitoring index T _z , the formula is as follows:

${T T}_{Z Z} = = \frac{{D D.}_{Z Z}}{{D D.}^{mod mod e e l l}}$

式中，D_z为第z个批次的监控量，D^model为监控控制限。In the formula, D _z is the monitoring amount of the zth batch, and D ^model is the monitoring control limit.

获得当前生产过程中某一批次的监控量和监控指标。Obtain the monitoring quantity and monitoring indicators of a certain batch in the current production process.

(5)在线监控(5) Online monitoring

获得当前生产过程中某一批次的监控量和监控指标后，即可进行比对判断，具体为：After obtaining the monitoring quantity and monitoring indicators of a certain batch in the current production process, comparison and judgment can be made, specifically:

当T_z≤1时，判断该批次过程合格，可进行实时更新；When T _z ≤ 1, it is judged that the batch process is qualified and can be updated in real time;

当T_z＞1时，判断批次过程不合格，此时发出警报，进行人为干涉处理。When T _z > 1, it is judged that the batch process is unqualified, and an alarm is issued at this time, and human intervention is performed.

本实施例中，对于第15批次过程D₁₅＝423.7022，从而，T₁₅＝0.94＜1，在线监控判断该批次过程合格。对于第88批次过程D₈₈＝615.7231，从而，T₈₈＝1.36＞1，在线监控判断该批次过程不合格。In this embodiment, for the 15th batch process D ₁₅ =423.7022, therefore, T ₁₅ =0.94<1, the online monitoring judges that the batch process is qualified. For the 88th batch process, D ₈₈ =615.7231, thus, T ₈₈ =1.36>1, and the online monitoring judged that the batch process was unqualified.

图3是本发明实施例提供的注塑成型过程曲线监控示意图，图中提供了第10批次，第15批次以及第88批次的系统压力与采样序号的关系图，由图可以明显看出，第15个批次过程数据曲线则与作为建模数据的第10个批次过程数据曲线接近，而第88个批次过程数据曲线则存在着明显的偏差，说明本实施例对于成型过程的监控判断是准确的。Fig. 3 is a schematic diagram of the injection molding process curve monitoring provided by the embodiment of the present invention, in which the relationship diagrams between the system pressure and the sampling number of the 10th batch, the 15th batch and the 88th batch are provided, which can be clearly seen from the figure , the 15th batch process data curve is close to the 10th batch process data curve as the modeling data, while the 88th batch process data curve has obvious deviations, which shows that the present embodiment has a strong influence on the molding process. The monitoring judgment is accurate.

(6)模型更新(6) Model update

对模型的更新，即为对监控模板和监控控制限的更新，对监控模板和监控控制限的更新，可以实时保证下一批次和上一批次的生产过程变量的一致性，可以很好的适应注塑成型过程的系统漂移。The update of the model is the update of the monitoring template and the monitoring control limit. The updating of the monitoring template and the monitoring control limit can ensure the consistency of the production process variables of the next batch and the previous batch in real time, which can be very good Systematic drift of adaptation to injection molding process.

当在线监控的注塑成型批次的过程正常时，更新现有监控模板和监控控制限。其中，When the process of the injection molding batch monitored online is normal, the existing monitoring template and monitoring control limits are updated. in,

监控模板的更新如下：The monitoring templates are updated as follows:

式中，λ为学习因子，可为任意非负数，SP^model为监控模板，SP_z为第z个批次的特征量向量。In the formula, λ is the learning factor, which can be any non-negative number, SP ^model is the monitoring template, and SP _z is the feature vector of the zth batch.

监控控制限更新如下：The monitoring control limits are updated as follows:

D^new-model＝(1-β)||SP^new-model||D ^new-model ＝(1-β)||SP ^new-model ||

式中，β为监控可靠度。In the formula, β is the monitoring reliability.

本实施例中，对于在线监控判断正常的批次过程，启用模型更新。模板更新时，设定学习因子式中B为当前建模数据中的批次个数。本实施例中，当在线监控预测第15个批次过程时，B＝5，从而监控可靠度β＝0.99，则In this embodiment, for the batch process judged to be normal by online monitoring, model update is enabled. When the template is updated, set the learning factor In the formula, B is the number of batches in the current modeling data. In this embodiment, when online monitoring predicts the 15th batch process, B=5, thus Monitoring reliability β=0.99, then

监控模板更新为：The monitoring template is updated to:

${SP SP}^{n no e e w w - - mod mod e e l l} = = ((11 - - \frac{11}{66})) \cdot \cdot {SP SP}^{mod mod e e l l} + + \frac{11}{66} \cdot \cdot {SP SP}_{1515},,$

监控控制限更新：Monitoring control limit update:

D^new-model＝(1-0.99)||SP^new-model||。D ^new-model = (1-0.99) ||SP ^new-model ||.

分别用SP^new-model、D^new-model代替SP^model、D^model即完成监控模型的更新。Replace SP ^model and D ^model with SP ^new-model and D ^new-model respectively to complete the update of the monitoring model.

本发明中涉及如下概念：特征量，特征量向量，特征量向量矩阵，其中特征量为一个过程变量在一个子阶段内的采样数据的统计量，包括最小值、最大值、平均值、方差、偏度、峰度、信息熵，还可能包括分位点、差分和积分。特征量向量为一个批次内的所有过程变量在所有子阶段内的特征量按列展开所得到的特征量向量，特征量向量矩阵为多个批次的特征向量按行展开得到的矩阵。The present invention involves the following concepts: feature quantity, feature quantity vector, feature quantity vector matrix, wherein the feature quantity is the statistical quantity of sampling data of a process variable in a sub-stage, including minimum value, maximum value, mean value, variance, Skewness, kurtosis, entropy, and possibly quantiles, differencing, and integrating. The feature quantity vector is the feature quantity vector obtained by expanding the feature quantities of all process variables in all sub-stages in a batch by column, and the feature quantity vector matrix is the matrix obtained by expanding the feature vectors of multiple batches by row.

本发明中，系统压力是指电机或者液压缸中作用在螺杆上以推动螺杆进行螺旋运动的力。模具温度为模具内壁的温度。料筒一般有多段，每段都测量温度，因此料筒温度具有多个。喷嘴是位于注塑机上的锥形结构，熔融塑料经该喷嘴从料筒注入到型腔并在型腔内成型。In the present invention, the system pressure refers to the force acting on the screw rod in the motor or hydraulic cylinder to push the screw rod to perform spiral motion. The mold temperature is the temperature of the inner wall of the mold. The barrel generally has multiple sections, and each section measures the temperature, so the barrel temperature has multiple. The nozzle is a conical structure located on the injection molding machine through which molten plastic is injected from the barrel into the cavity and molded in the cavity.

本发明中，各个子阶段按照时间顺序连串起来组成一个周期，一个周期即是一个批次。根据注塑成型过程内在机理，可选择监控一个或者多个子阶段的过程数据。In the present invention, each sub-stage is connected in sequence according to time to form a cycle, and a cycle is a batch. According to the intrinsic mechanism of the injection molding process, one or more sub-stage process data can be selected to be monitored.

本发明实施例中仅仅采用系统压力和螺杆位置为例进行了详细说明，对于螺杆转速、螺杆行进速度、料筒温度、喷嘴温度、模具温度、塑化阶段的塑化压力、型腔压力以及喷嘴压力等等过程变量，均可采用如上所述的方法进行监控。在以上的过程变量中，系统压力和螺杆位置是非常关键的过程变量，需要全程监控。在注射阶段的充模速度速度，保压阶段的保压压力，射退阶段的螺杆位置也是需要被重点关注的过程变量。In the embodiment of the present invention, only the system pressure and screw position are used as examples to describe in detail. For the screw speed, screw travel speed, barrel temperature, nozzle temperature, mold temperature, plasticizing pressure in the plasticizing stage, cavity pressure and nozzle Process variables such as pressure can be monitored by the method described above. Among the above process variables, system pressure and screw position are very critical process variables that need to be monitored throughout the process. The filling speed in the injection stage, the holding pressure in the holding stage, and the screw position in the ejection stage are also process variables that need to be focused on.

本领域的技术人员容易理解，以上所述仅为本发明的较佳实施例而已，并不用以限制本发明，凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等，均应包含在本发明的保护范围之内。It is easy for those skilled in the art to understand that the above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention, All should be included within the protection scope of the present invention.

Claims

1. an injection molding process online monitoring method, for monitoring the injection molding machine production process, is characterized in that, comprises the steps:

S1: Collect the j-th process variable x _i,m,j of the m-th sub-stage of the i-th batch in the injection molding process, where, i=1,2,…,I; j=1,2,…, J; m=1,2,...,M, I is the total number of batches, J is the total number of process variables, M is the total number of sub-stages, and the sub-stages include at least the injection stage, the pressure holding stage, the plasticizing stage and the ejection stage , the process variables include at least system pressure and screw stroke;

S2: First calculate the feature quantity and feature quantity vector of x _i,m,j mentioned in step S1, the feature quantity includes the minimum value σ _i,m,j,1 and the maximum value σ _i,m,j of the process variable _,2 , mean value μ _i,m,j , variance ε _i,m,j , skewness γ _i,m,j , kurtosis κ _i,m,j , information entropy The feature quantity vector is as follows:

In each of the above feature quantities and the feature quantity vector, i=1,2,...,I; j=1,2,...,J; m=1,2,...,M, and the total number of process variables J≥2, The total number of sub-stages M≥4;

Then calculate and obtain the feature quantity vector SP _i of the i-th batch, and the SP _i is as follows:

SP _i =vec[SP _i,m,j ]

＝[SP _i,1,1 ,SP _i,1,2 ,...,SP _i,1,J ,SP _i,2,1 ,SP _i,2,2 ,...,SP _i,2,J ,..., SP _i,M,1 ,SP _i,M,2 ,…,SP _i,M,J ] ^T

Among them, vec[ ] means to arrange the data into a column vector form;

S3: Calculate and obtain the monitoring template SP ^model of B qualified batches and the monitoring control limit D ^model of B qualified batches, which are as follows:

D ^model ＝(1-β)||SP ^model ||

In the above formula, ||·|| is the vector 2 norm calculation, β is the monitoring reliability, β can take any value in the range (0, 1), B is the number of batches and B is less than 1, is the feature vector of the bth batch in a total of B batches;

S4: Calculate the monitoring quantity D _z and monitoring index T _Z of the z-th batch in the current production process, respectively as follows:

D _z ＝||SP _z -SP ^model ||

In the above two formulas, SP _z is the feature quantity vector of the zth batch, and SP ^model is the monitoring template of B qualified batches and the monitoring control limit D ^model of B qualified batches;

S5: Carry out online monitoring and judgment. When T _z ≤ 1, it is judged that the production process of the z-th batch is qualified; when T _z > 1, it is judged that the production process of the z-th batch is abnormal, and the online injection molding process is realized. monitor.

2. A kind of injection molding process online monitoring method as claimed in claim 1, is characterized in that, also comprises step S6, and described step S6 is to the monitoring template SP ^model of B qualified batches and B qualified batches The monitoring control limit D ^model is updated, and the monitoring template SP model of B qualified batches is replaced with the new monitoring template SP ^new- ^model . The new monitoring template SP ^new-model is as follows:

SP ^new-model = (1-λ) SP ^model + λ SP _z

In the formula, SP _z is the feature vector of the zth batch, SP ^model is the monitoring template of B qualified batches, and λ is the learning factor, which is any non-negative number;

Replace the monitoring control limit D model of B qualified batches with the new monitoring control limit D ^new - ^model . The new monitoring control limit D ^new-model is as follows:

D ^new-model ＝(1-β)||SP ^new-model ||

In the formula, SP ^new-model is the new monitoring template, and β is the monitoring reliability.

3. A method for online monitoring of injection molding process according to claim 1, characterized in that the minimum value σ _i,m,j,1 , the maximum value σ _i,m,j,2 , the average value of the process variable Value μ _i,m,j , variance ε _i,m,j , skewness γ _i,m,j , kurtosis κ _i,m,j , information entropy Calculated using the following formulas:

σ _i,m,j,1 = min[ _xi,m,j (α)]

σ _i,m,j,2 = max[ _xi,m,j (α)]

In the formula, α=1,2,...,K _i,m,j , where K _i,m,j is the length of x _i,m,j .

4. The online monitoring method of injection molding process according to claim 1, characterized in that the process variable is collected by a sensor attached to the injection molding machine.

5. A method for online monitoring of injection molding process according to claim 1, characterized in that said process variables also include screw speed, screw travel speed, barrel temperature, nozzle temperature, mold temperature, plasticizing stage Vulcanization pressure, cavity pressure, nozzle pressure, filling speed in the injection stage and holding pressure in the holding stage.

6 . The online monitoring method of injection molding process according to claim 1 , wherein the feature quantity further includes a quantile point, a difference value and an integral. 7 .

7. An online monitoring method for injection molding process according to claim 1, characterized in that, said sub-stages further include a mold closing stage, a cooling stage and a mold opening stage.