Zero-defect quality control method and device for injection molding part production

Abstract

The invention is applicable to the technical field of injection molding control, and provides a zero-defect quality control method and a zero-defect quality control device for production of injection molded parts, and the method is applied to a zero-defect quality control system for production of injection molded parts. The system comprises an information acquisition module, a data processing module, a model construction module, a model training module, an injection molding module, a wireless communication module, a memory, an alarm, a processing center and an intelligent control terminal. The information acquisition module, the data processing module, the model construction module, the model training module, the injection molding module, the wireless communication module, the memory and the alarm are respectively connected with the processing center; the intelligent mobile phone, the tablet personal computer or the intelligent remote controller is in wireless network connection with the wireless communication module in the range of the wireless network or the Internet; the invention further provides a zero-defect quality control device for injection molding part production. According to the invention, process parameters can be automatically determined and adjusted, and automatic detection is realized.

Description

Technical Field

[0001] The present invention relates to the field of injection molding control technology, and in particular to a method and device for zero-defect quality control of injection molded parts production in the field of artificial intelligence. Background Art

[0002] Most of the existing injection molding companies use ERP management systems to obtain a summary of injection molding production data so that managers can discover various abnormal problems in injection molding and take measures to improve them. However, human decision-making sometimes leads to molding abnormalities due to misjudgment, unreasonable process parameter design, human input parameter errors, or human detection errors, resulting in problems such as "repeated adjustment of molding process, repeated repair of molds, and human detection errors", resulting in low injection molding production efficiency, extended mold modification time, and increased injection molding costs. Summary of the invention

[0003] In order to solve the deficiencies in the prior art, the purpose of the present invention is to provide a method and device for zero-defect quality control of injection molded parts production. By setting an information acquisition module, a data processing module, a model building module, a model training module, and an injection molding module, the process parameters can be automatically determined, the process parameters can be automatically adjusted, and automatic detection can be performed, thereby realizing automated production and detection of injection molding, avoiding the problems of "repeated adjustment of molding process, repeated repair of molds, and human detection errors", improving injection molding production efficiency, and reducing high labor costs.

[0004] In order to achieve the above purpose, the technical solution adopted by the present invention is: A method for zero-defect quality control of injection molded parts production is applied to a zero-defect quality control system for injection molded parts production, the system comprising an information acquisition module, a data processing module, a model building module, a model training module, an injection molding module, a wireless communication module, a memory, an alarm, a processing center, and an intelligent mobile terminal; the information acquisition module, the data processing module, the model building module, the model training module, the injection molding module, the wireless communication module, the memory, and the alarm are respectively connected to the processing center; the intelligent mobile terminal comprises a smart phone, a tablet computer or an intelligent remote control, which are respectively connected to the wireless network of the wireless communication module within the range of a wireless network or the Internet; The wireless communication module is provided with a wireless network unit, which is responsible for sending and receiving wireless signals and automatically networking with the intelligent control terminal within an effective network range; The alarm device compares the actual evaluation score of the model test with the evaluation score standard of the injection molded part abnormal identification training model test stored in the memory, and automatically sounds an alarm if the standard is not met and notifies to continue training until the standard is met; and compares the various quality information of the actual injection molded part with the quality standard of the injection molded part stored in the memory, and automatically sounds an alarm if the standard is not met and notifies to rework; The memory is responsible for storing information of the information acquisition module, data processing module, model building module, model training module, injection molding module, wireless communication module, and alarm, as well as the storage of injection molded parts abnormality recognition training model test evaluation sub-standards and injection molded parts quality standards; The processing center is responsible for information collection module, data processing module, model building module, model training module, injection molding module, wireless communication module, alarm, and information transmission of memory. It is the hub center of the system and compares the actual evaluation score of the model test with the injection molding abnormality recognition training model test evaluation score standard stored in the memory: if it meets the standard, it is determined as a molding process parameter to notify the injection molding production; if it does not meet the standard, it is transmitted to the alarm and notified to continue training; and compares the various quality information of the actual injection molding parts with the quality standard of the injection molding parts stored in the memory: if it meets the standard, production continues; if it does not meet the standard, it is transmitted to the alarm and notified to rework; The information collection module includes a mold information unit, a molding information unit, and a data mining unit, which is responsible for collecting various characteristic information and historical molding process data of the injection mold to be molded and processing them for subsequent model construction, and passing them to the data processing module; The data processing module includes a standard data unit, a smoothing processing unit, a filtering and denoising unit, and a data analysis unit, which are responsible for standardizing, smoothing, filtering and denoising, and classifying and identifying the mined data for subsequent model construction and passing it to the model construction module; The model building module includes a network structure unit, a forward transmission unit, a back-propagation unit, a linear transformation unit, a ReLU function unit, and an anti-overfitting unit, which is responsible for building an injection molding abnormality type prediction model and its required training set and test set through a convolutional neural network algorithm according to the identified abnormal pattern, and passing them to the model training module; The model training module includes a normalization processing unit, a convolution output unit, a pooling calculation unit, a gradient optimization unit, a training learning unit, and a test evaluation unit. The model is trained according to the collected data model and the model parameters are adjusted to optimize the model performance to determine the process parameters, ensure the stability of the molding quality, and pass them to the injection molding module; The injection molding module includes a molding application unit, an appearance inspection unit, and a performance testing unit. It performs injection molding according to the process parameters predicted by the trained model and obtains various quality information of the injection molded parts, and transmits it to the processing center.

[0005] The present invention provides a method for zero-defect quality control of injection molded parts production, comprising the following steps: S10. During production, the information collection module collects various characteristic information and historical molding process data of the injection mold to be formed and processes them for subsequent model construction, and transmits them to the data processing module; S20, the data processing module standardizes, smoothes, filters, removes noise, and classifies the mined data for subsequent model construction, and passes it to the model construction module; S30, the model building module builds an injection molding abnormality type prediction model and its required training set and test set according to the identified abnormal pattern through a convolutional neural network algorithm, and passes it to the model training module; S40, the model training module trains the model according to the collected data model and adjusts the model parameters to optimize the model performance to determine the process parameters, ensure the stability of the molding quality, and pass it to the injection molding module; S50, the injection molding module performs injection molding according to the process parameters predicted by the trained model and obtains various quality information of the injection molded parts, and transmits it to the processing center; S60, the processing center compares the quality information of the actual injection molded parts with the quality standards of the injection molded parts stored in the memory: if the standards are met, production continues; if the standards are not met, the information is transmitted to the alarm and a notice is given to rework.

[0006] Further, the step S10 comprises the following steps: S11, the mold information unit obtains characteristic information of various injection molds to be molded through the intelligent sensor and data collector set on the injection molding machine, so as to collect subsequent molding parameters and transmit them to the molding information unit; S12, the molding information unit obtains various molding data of the corresponding injection mold by networking and matching with the industry database according to the collected mold feature information, so as to facilitate subsequent data processing and transmit it to the data mining unit; S13, the data mining unit calculates the formula " , y' is the predicted output, w is the perceptron model parameter, x is the linear perceptron model attribute, and yi is the actual output" to obtain the cleaned and mined data.

[0007] Further, the step S20 comprises the following steps: S21, the standard data unit integrates, removes duplication, removes differences, unifies, and pre-processes the mined data to ensure data standardization for subsequent smoothing processing, and transmits it to the smoothing processing unit; S22, the smoothing processing unit obtains the smoothed data after mining according to the exponential smoothing calculation formula "St=α*yt-1+(1-α)St-1, St is the smoothing value of the tth period, yt-1 is the actual value of the t-1th period, and α is the smoothing coefficient" to ensure the integrity of the data for subsequent filtering and denoising, and pass it to the filtering and denoising unit; S23, the filtering and denoising unit obtains the filtered and denoised data according to the first-order low-pass filtering calculation formula "y(t)=y(t-1)+K*[u(t)-y(t-1)], y(t) is the output signal at the current moment, K is the signal balance coefficient (0-1), u(t) is the input signal at the current moment, and y(t-1) is the output signal at the previous moment", and transmits it to the data analysis unit; S24. The data analysis unit uses deep learning algorithms to conduct in-depth analysis of the collected data and identify six potential abnormal patterns of injection molded parts, optimizing the production process for subsequent model construction.

[0008] Further, the step S30 comprises the following steps: S31, the network structure unit identifies the abnormal pattern classification label of the injection molded part and establishes the network structure of the input layer, convolution layer, pooling layer, full connection layer, and output layer, and transmits it to the forward transmission unit; S32, the forward transfer unit calculates the convolutional layer forward transfer formula "a'=σ( Wlm*Xm+bl), a' is the output of the lth layer, σ is the activation function, Wlm is the convolution kernel of the lth layer, Xm is the input feature map, bl is the bias term" to obtain the convolution layer output and pass it to the back propagation unit; S33, the back propagation unit obtains the output of the convolution layer and the fully connected layer according to the back propagation calculation formula "dz[l]=da[l]*g[l]′(z[l]), da[l] is the input data, da[l-1], dw[l], db[l] is the output data, g[l]′ is the activation function sigmoid derivative", and passes it to the linear transformation unit; S34, the linear transformation unit obtains the linear transformation output according to the linear transformation calculation formula "y=W*X+b, X is the input data, y is the output data, W is the weight matrix, and b is the bias vector", and passes it to the ReLU function unit; S35, the ReLU function unit obtains the output of the Tanh function according to the ReLU function calculation formula "f(x)=max(0,x), f(x) is the output of the ReLU function, and x is the input of the ReLU function", and passes it to the anti-overfitting unit; S36. The anti-overfitting unit obtains the TDropout output according to the Dropout calculation formula "y=f(W*X+b), y is the neural network output, f is the activation function, W is the weight matrix, X is the input data, and b is the bias vector)".

[0009] Further, the step S40 comprises the following steps: S41, the normalization processing unit obtains the output of Batch Norm according to the Batch Norm calculation formula "h=Φ(BN(Wx+b)), h is the output of the fully connected layer, Φ is the activation function, BN is the batch normalization operator, W is the weight parameter, x is the input of the fully connected layer, and b is the bias parameter", and passes it to the convolution output unit; S42, the convolution output unit calculates the convolution output according to the convolution output formula "O(n,k,ox,oy)= I(n,c,ix,iy)*W(k,c,r,s), O(n,k,ox,oy) convolution output data, C is the convolution channel, R is the convolution kernel height, S is the convolution kernel width, I(n,c,ix,iy), W(k,c,r,s) are respectively the convolution input data "to obtain the convolution output data, and pass it to the pooling calculation unit; S43, the pooling calculation unit calculates the maximum pooling algorithm according to the formula "M(i,j)= A(i*s+u,j*s+v), M(i,j) is the (i,j)th element of the output feature map, u and v are indexes that vary in the range [0,f-1], and s is the step size. The spatial dimension of the feature map is obtained and passed to the gradient optimization unit; S44, the gradient optimization unit obtains the gradient descent data according to the gradient descent method calculation formula "Xn+1=Xn-α*▽f(Xn), Xn is the nth iteration parameter value, α is the learning rate or step size, ▽f(Xn) is the gradient of the function f at Xn", and passes it to the training learning unit; S45, the training learning unit determines the training model according to the gradient descent data, and divides it into a training set, a validation set, and a test set, performs model training, and passes it to the test evaluation unit; S46. The test evaluation unit obtains the F1 comprehensive evaluation score according to the training model comprehensive evaluation score calculation formula "F1=2*(Ac*Re) / (Ac+Re), Ac=(TP+TN) / (TP+FP+FN+TN), Re=TP / (TP+FN), F1 is the training model evaluation score, Ac is the precision, Re is the recall, TP is the number of samples that are actually positive and correctly predicted to be positive, FP is the number of samples that are actually negative but incorrectly predicted to be positive, FN is the number of samples that are actually positive but incorrectly predicted to be negative, TN is the number of samples that are actually negative and correctly predicted to be negative" and transmits it to the processing center; S47. The processing center compares the actual evaluation score of the model test with the evaluation score standard of the injection molded part abnormal recognition training model test stored in the memory: if it meets the standard, it is determined as a molding process parameter to notify the injection molding production; if it does not meet the standard, it is transmitted to the alarm and notified to continue training until it meets the standard.

[0010] Further, the step S50 comprises the following steps: S51, the molding application unit automatically inputs the system according to the determined molding process parameters and debugs them in place for injection molding production, ensuring that the quality information of the actual injection molded parts is consistent with the model, and transmits it to the appearance inspection unit; S52, the appearance inspection unit obtains the appearance image of the molded injection molded part through a set high-definition camera or image sensor and converts it into recognizable various quality information of the injection molded part appearance, and transmits it to the performance inspection unit; S53. The performance testing unit obtains the dimension data, physical performance test data, and chemical performance test data of the injection molded parts with normal appearance through the injection molded parts quality inspection equipment to meet the customer's needs for injection molded parts.

[0011] A zero-defect quality control system for injection molded parts production provided by the present invention also includes a computer-readable storage medium containing a memory; the memory stores a computer program, and when the above functional modules execute the computer program, the steps of the zero-defect quality control method for injection molded parts production described in any one of the above are implemented; the computer-readable storage medium stores a computer program, and when the computer program is executed by each functional module, the steps of the zero-defect quality control method for injection molded parts production described in any one of the above are implemented.

[0012] The present invention also provides a zero-defect quality control device for injection molded parts production, which is implemented by the above-mentioned zero-defect quality control method for injection molded parts production.

[0013] The beneficial effects of the present invention compared with the prior art are as follows: By setting up information collection module, data processing module, model building module, model training module and injection molding module, the process parameters can be automatically determined, automatically adjusted and automatically detected, thereby realizing automated production and detection of injection molding, avoiding the problems of "repeated adjustment of molding process, repeated repair of mold and human detection errors", improving injection molding production efficiency and reducing high labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or exemplary technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0015] Figure 1 It is a schematic diagram of the system module of the present invention; Figure 2 This is a schematic diagram of the information collection module of the present invention; Figure 3 It is a schematic diagram of a data processing module of the present invention; Figure 4 It is a schematic diagram of the model construction module of the present invention; Figure 5 This is a schematic diagram of a model training module of the present invention; Figure 6 It is a schematic diagram of the injection molding module of the present invention; Figure 7 It is a schematic diagram of the method flow control program of the present invention; Figure 8 This is a schematic diagram of step S10 in the method flow of the present invention; Fig. 9 This is a schematic diagram of step S20 in the method flow of the present invention; Fig.10 This is a schematic diagram of step S30 in the method flow of the present invention; Fig.11 This is a schematic diagram of step S40 in the method flow of the present invention; Fig.12 It is a program schematic diagram of step S50 in the method flow of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments: In order to make the technical problems, technical solutions and beneficial effects to be solved by this application more clearly understood, the application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0018] It should be noted that when a module is referred to as being "disposed on" another module, it can be directly on the other module or indirectly on the other module. When a module is referred to as being "connected to" another module, it can be directly connected to the other module or indirectly connected to the other module.

[0019] In the description of this application, "multiple" means two or more, unless otherwise clearly defined. "Several" means one or more, unless otherwise clearly defined.

[0020] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "connected" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0021] See also Figure 1 As shown, the present invention provides a method for zero-defect quality control of injection molded parts production, which is applied to a zero-defect quality control system for injection molded parts production, wherein the system comprises an information acquisition module, a data processing module, a model building module, a model training module, an injection molding module, a wireless communication module, a memory, an alarm, a processing center, and an intelligent mobile terminal; the information acquisition module, the data processing module, the model building module, the model training module, the injection molding module, the wireless communication module, the memory, and the alarm are respectively connected to the processing center; the intelligent mobile terminal comprises a smart phone, a tablet computer or a smart remote control, which are respectively connected to the wireless communication module wireless network within the range of a wireless network or the Internet, wherein the wireless network comprises networks such as the Internet of Things and Wi-Fi.

[0022] The wireless communication module is provided with a wireless network unit, which is responsible for sending and receiving wireless signals and automatically networking with the intelligent control terminal within an effective network range.

[0023] The alarm compares the actual evaluation score of the model test with the injection molded part abnormality recognition training model test evaluation score standard stored in the memory. If the standard is not met, it will automatically sound an alarm and notify to continue training until the standard is met; and compares the various quality information of the actual injection molded parts with the injection molded part quality standard stored in the memory. If the standard is not met, it will automatically sound an alarm and notify to rework.

[0024] The memory is responsible for storing information of the information acquisition module, the data processing module, the model construction module, the model training module, the injection molding module, the wireless communication module, and the alarm, as well as storing injection molded part abnormality recognition training model test evaluation points and injection molded part quality standards.

[0025] The processing center is responsible for information collection module, data processing module, model building module, model training module, injection molding module, wireless communication module, alarm, and information transmission of memory. It is the hub center of the system and compares the actual evaluation score of the model test with the injection molded parts abnormal recognition training model test evaluation score standard stored in the memory: if it meets the standard, it is determined as a molding process parameter to notify the injection molding production; if it does not meet the standard, it is passed to the alarm and notified to continue training; and compares the various quality information of the actual injection molded parts with the injection molded parts quality standard stored in the memory: if it meets the standard, production continues; if it does not meet the standard, it is passed to the alarm and notified to rework.

[0026] See also Figure 2 As shown, the information acquisition module includes a mold information unit, a molding information unit, and a data mining unit, which is responsible for collecting and processing various feature information and historical molding process data of the injection mold to be molded, so as to facilitate subsequent model construction and pass it to the data processing module.

[0027] Furthermore, the mold information unit obtains characteristic information of various injection molds to be molded through the intelligent sensor and data collector set on the injection molding machine, and transmits it to the molding information unit; the molding information unit obtains various molding data of the corresponding injection mold through networking matching with the industry database according to the collected mold characteristic information, and transmits it to the data mining unit; the data mining unit calculates the formula "y'=sign (wd*xd+wd-1*xd-1)” to obtain the mining molding data after cleaning.

[0028] See also Figure 3 As shown, the data processing module includes a standard data unit, a smoothing processing unit, a filtering and denoising unit, and a data analysis unit, which are responsible for standardizing, smoothing, filtering and denoising, and classifying and identifying the mined data for subsequent model construction and passing it to the model construction module.

[0029] Furthermore, the standard data unit integrates, removes duplications, eliminates differences, unifies, and standardizes the mined data, and passes it to the smoothing processing unit; the smoothing processing unit obtains the mined smoothed molding data according to the exponential smoothing calculation formula "St=α*yt-1+(1-α)St-1", and passes it to the filtering and denoising unit; the filtering and denoising unit obtains the filtered and denoised molding data according to the first-order low-pass filtering calculation formula "y(t)=y(t-1)+K*[u(t)-y(t-1)]", and passes it to the data analysis unit; the data analysis unit performs in-depth analysis of the collected data through a deep learning algorithm and identifies six potential abnormal modes of injection molded parts for subsequent model construction.

[0030] See also Figure 4 As shown, the model building module includes a network structure unit, a forward transmission unit, a back-propagation unit, a linear transformation unit, a ReLU function unit, and an anti-overfitting unit, which is responsible for building an injection molding abnormality type prediction model and its required training set and test set through a convolutional neural network algorithm according to the identified abnormal pattern, and passing it to the model training module.

[0031] Further, the network structure unit sets up a network structure of input layer, convolution layer, pooling layer, full connection layer and output layer according to the identified abnormal pattern classification label of the injection molded part, and transmits it to the forward transmission unit; the forward transmission unit calculates the forward transmission formula of the convolution layer according to the forward transmission formula "a'=σ( Wlm*Xm+bl)" to obtain the convolution layer output and pass it to the back propagation unit; the back propagation unit obtains the output of the convolution layer and the fully connected layer according to the back propagation calculation formula "dz[l]=da[l]* g[l]′(z[l])" and passes it to the linear transformation unit; the linear transformation unit obtains the linear transformation output according to the linear transformation calculation formula "y=W*X+b" and passes it to the ReLU function unit; the ReLU function unit obtains the output of the Tanh function according to the ReLU function calculation formula "f(x)=max(0,x)", and passes it to the anti-overfitting unit; the anti-overfitting unit obtains the TDropout output according to the Dropout calculation formula "y=f(W*X+b)".

[0032] See also Figure 5 As shown, the model training module includes a normalization processing unit, a convolution output unit, a pooling calculation unit, a gradient optimization unit, a training learning unit, and a test evaluation unit. The model is trained according to the collected data and the model parameters are adjusted to optimize the model performance to determine the process parameters, ensure the stability of the molding quality, and pass them to the injection molding module.

[0033] Further, the normalization processing unit obtains the output of Batch Norm according to the Batch Norm calculation formula “h=Φ(BN(Wx+b))” and passes it to the convolution output unit; the convolution output unit calculates the output of Batch Norm according to the convolution output calculation formula “O(n, k, ox, oy)= I(n, c, ix, iy)*W(k, c, r, s)” obtains the convolution output data and passes it to the pooling calculation unit; the pooling calculation unit calculates the maximum pooling algorithm according to the formula “M(i, j)= A(i*s+u,j*s+v)" obtains the spatial dimension of the feature map and passes it to the gradient optimization unit; the gradient optimization unit obtains the gradient descent data according to the gradient descent method calculation formula "Xn+1=Xn-α*▽f(Xn)", and passes it to the training learning unit; the training learning unit determines the training model according to the gradient descent data, and divides it into a training set, a validation set, and a test set to perform model training, and passes it to the test evaluation unit; the test evaluation unit obtains the F1 comprehensive evaluation score according to the training model comprehensive evaluation score calculation formula "F1=2*(Ac*Re) / (Ac+Re), Ac=(TP+TN) / (TP+FP+FN+TN), Re=TP / (TP+FN)", and passes it to the processing center.

[0034] See also Figure 6 As shown, the injection molding module includes a molding application unit, an appearance inspection unit, and a performance testing unit. Injection molding is performed according to the process parameters predicted by the trained model, and various quality information of the injection molded parts is obtained and transmitted to the processing center.

[0035] Furthermore, the molding application unit automatically inputs the determined molding process parameters into the system and performs injection molding production after debugging, so as to ensure that the quality information of the appearance of the actual injection molded parts is consistent with the model, and transmits it to the appearance detection unit; the appearance detection unit obtains the appearance image of the injection molded parts after molding through the set high-definition camera or image sensor and converts it into recognizable quality information of the injection molded parts appearance, and transmits it to the performance detection unit; the performance test unit obtains the dimensional data, physical property test data, and chemical property test data of the injection molded parts with no abnormal appearance through the injection molded parts quality detection equipment, so as to meet the customer's demand for injection molded parts.

[0036] System operation working principle: During production, the information collection module collects various characteristic information and historical molding process data of the injection mold to be molded and processes them for subsequent model construction, and passes them to the data processing module; then the data processing module standardizes, smoothes, filters, de-noises, and classifies the mined data for subsequent model construction, and passes them to the model construction module; the model construction module constructs an injection molding abnormality type prediction model and its required training set and test set through the convolutional neural network algorithm according to the identified abnormal pattern, and passes them to the model training module; then the model training module trains the model according to the collected data model and adjusts the model parameters to optimize the model performance to determine the process parameters, ensure the stability of molding quality, and pass them to the injection molding module; the injection molding module performs injection molding according to the process parameters predicted by the trained model and obtains various quality information of the injection molded parts, and passes it to the processing center; the control processing center compares the various quality information of the actual injection molded parts with the quality standards of the injection molded parts stored in the memory: if the standards are met, production will continue; if the standards are not met, the alarm will be passed to the alarm and a notice will be given to rework.

[0037] When workshop managers or managers are away from home, they can use smart phones or tablets to automatically connect to the wireless network or the Internet through the wireless communication module, and realize the combination of smart terminals, the Internet and artificial intelligence technology to achieve remote monitoring, automatic identification and intelligent scheduling functions. Through the intelligent system, the production status of injection molded parts can be monitored remotely in real time through mobile phone APP or smart terminals, thereby improving molding efficiency and reducing injection molding costs.

[0038] See also Figure 7 As shown, the present invention provides a method for zero-defect quality control of injection molded parts production, comprising the following steps: S10. During production, the information collection module collects various characteristic information and historical molding process data of the injection mold to be formed and processes them for subsequent model construction, and transmits them to the data processing module; See also Figure 8 As shown, the step S10 includes the following steps: S11, the mold information unit obtains characteristic information of various injection molds to be molded through the intelligent sensor and data collector set on the injection molding machine, so as to collect subsequent molding parameters and transmit them to the molding information unit; It is further explained that various characteristic information of the injection mold that has been installed and prepared for injection molding is obtained through intelligent sensors and data collectors, including basic parameters such as mold size, cavity shape, internal structure, water transport layout, mold material, as well as injection molded part information, molding requirements, machine model, etc.; wherein, the data collector obtains the pre-stored mold design information in the system memory by networking with the system, and automatically screens the corresponding mold design information for collection, thereby obtaining the required various characteristic information of the corresponding injection mold.

[0039] S12, the molding information unit obtains various molding data of the corresponding injection mold by networking and matching with the industry database according to the collected mold feature information, so as to facilitate subsequent data processing and transmit it to the data mining unit; It is further explained that the industry database includes a mold injection molding database independently established by the enterprise based on the mold information and its raw material data, process parameter data, injection molded part quality data, and machine status data recorded in each injection molding of different molds, as well as a mold injection molding database shared by other peer companies in the industry at home and abroad, and network shared data information is acquired in real time to continuously enrich its accumulated industry data information; the various molding data of the injection mold include: raw material data such as raw material batch, raw material brand and model, raw material temperature, injection molding machine parameter data such as injection speed and pressure, holding time and pressure, cooling time and temperature, cycle, start-up time, running time, injection molding machine temperature and pressure, fault and alarm information, output, efficiency, energy consumption and other injection molding machine status data, product size and weight, appearance and surface quality, mechanical properties, yield rate and other product quality data, so as to form a comprehensive and detailed data collection system in the injection molding process, and provide accurate data for subsequent data intelligence.

[0040] S13, the data mining unit calculates the formula "y'=sign(wd*xd+wd-1*xd-1), E(w)=1 / 2* (yi-y')², y' is the predicted output, w is the perceptron model parameter, x is the linear perceptron model attribute, and yi is the actual output" to obtain the cleaned and mined data.

[0041] It is further explained that before data mining, the injection molding equipment data retrieved from the system database is first cleaned to remove the data belonging to outliers, and the cleaned injection molding equipment data is obtained to remove outliers or erroneous data to ensure the accuracy and reliability of the data; and the calculation formula of the artificial neural network method in the pre-trained data mining algorithm is used, "y'=sign(wd*xd+wd-1*xd-1), E(w)=1 / 2* (yi-y')²" to obtain the mining data in the injection molding equipment data after cleaning, where "y'=sign(wd*xd+wd-1*xd-1)" is inferred through the output formula of the perceptron model "y'=sign(wd*xd+wd-1*xd-1+…+w2*x2+w1x1-t)", so that useful information and patterns can be extracted from the cleaned data to support decision-making or further analysis, w0=-t, x0=1, so that useful data can be effectively obtained from massive injection molding equipment data, the utilization efficiency of data can be improved, and data support can be provided for the optimization and improvement of the injection molding process.

[0042] S20, the data processing module standardizes, smoothes, filters, removes noise, and classifies the mined data for subsequent model construction, and passes it to the model construction module; See also Fig. 9 As shown, the step S20 includes the following steps: S21, the standard data unit integrates, removes duplication, removes differences, unifies, and pre-processes the mined data to ensure data standardization for subsequent smoothing processing, and transmits it to the smoothing processing unit; It is further explained that the scattered data after mining are integrated according to classification to ensure its centralization; the incomplete data after integration are spliced ​​to ensure its integrity; the repeated and overlapping data are cleared to remove the repeated or overlapping data; the data points beyond the reasonable range are processed to remove outliers; the data are normalized, standardized, and encoded category variables are preprocessed to convert the original data into a format suitable for neural network processing, and the image data is resized, cropped, rotated, etc.; after encoding, the number and diversity of training data are increased through rotation, scaling, flipping, color conversion, etc. to improve the generalization ability of the model; S22, the smoothing processing unit obtains the smoothed data after mining according to the exponential smoothing calculation formula "St=α*yt-1+(1-α)St-1, St is the smoothing value of the tth period, yt-1 is the actual value of the t-1th period, and α is the smoothing coefficient" to ensure the integrity of the data for subsequent filtering and denoising, and pass it to the filtering and denoising unit; It is further explained that the actual value yt-1 of the t-1 period is obtained by the sensor in a timely manner, and the data at different time points are different before the model training due to the instability of each process; the smoothing coefficient α is a known parameter in the industry; the normalized and smoothed data is obtained according to the smoothing calculation formula "St=α*yt-1+(1-α)St-1", which determines the relative importance of new data and old data, and helps to reduce the complexity of subsequent filtering processing and improve the filtering effect.

[0043] S23, the filtering and denoising unit obtains the denoised data according to the first-order low-pass filtering calculation formula "y(t)=y(t-1)+K*[u(t)-y(t-1)], y(t) is the output signal at the current moment, K is the signal balance coefficient (0-1), u(t) is the input signal at the current moment, and y(t-1) is the output signal at the previous moment", and transmits it to the data analysis unit; Further explanation, the first-order filtering calculation formula can also be used "y(t)=K*u(t)+(1-K)*y(t-1)" and the calculated data is the same. The signal balance coefficient K is the balance relationship between the input signal and the output signal at the previous moment. When K=1, there is no filtering effect. When K<1, the input signal is low-pass filtered. The low-pass filter allows low-frequency signals to pass through and suppresses high-frequency signals. Since high-frequency signals often represent noise in injection molding, low-pass filtering can effectively remove noise and provide a basis for subsequent analysis, control or decision-making. After the filtering process is completed, the filtering effect is evaluated and the data before and after filtering are compared. If the filtering effect is not ideal, the filtering algorithm or filtering parameters are adjusted, and the filtering process is performed again until the effect meets the requirements.

[0044] S24. The data analysis unit uses deep learning algorithms to conduct in-depth analysis of the collected data and identify six potential abnormal patterns of injection molded parts, optimizing the production process for subsequent model construction.

[0045] Further explanation: through data analysis, a variety of potential injection molding abnormal patterns can be identified, including the abnormal categories of injection molded parts, causes, solutions, and process parameters, wherein the abnormal categories of injection molded parts include but are not limited to insufficient filling, overflow, pores, ripples / silver stripes, surface halos, fusion lines / bubbles, etc.; the causes include: the volume of the injection molded parts is too large, the runner or gate is too small, the nozzle temperature is low, the material temperature or injection pressure is low, the clamping force is insufficient, the mold problem, the material temperature is too high or the material supply is excessive, the material is not fully dried or the melt contains volatiles or air, poor material flow, low mold temperature or poor exhaust, too much lubricant or too much demolding material, improper material temperature and gate design or too many volatile components in the material, machine failure, etc.; the solutions include: adjusting mold design, increasing material temperature and injection pressure, increasing clamping force, cleaning impurities on the mold surface, adjusting material supply, fully drying the material, adjusting injection pressure and speed, adjusting mold temperature and improving exhaust design, by reducing the use of lubricant or adjusting the amount of demolding material used, by adjusting material temperature and optimizing mold design, installing sensors to monitor the machine.

[0046] S30, the model building module builds an injection molding abnormality type prediction model and its required training set and test set according to the identified abnormal pattern through a convolutional neural network algorithm, and passes it to the model training module; See also Fig.10 As shown, the step S30 includes the following steps: S31, the network structure unit identifies the abnormal pattern classification label of the injection molded part and establishes the network structure of the input layer, convolution layer, pooling layer, full connection layer, and output layer, and transmits it to the forward transmission unit; It is further explained that the data set is passed through the convolutional neural network deep learning method in a 9:1 relationship to establish the training set and test set required for the injection molded parts abnormal pattern prediction model training, thereby establishing a deep convolutional neural network structure with a depth of 10, including 3 convolutional layers, 3 maximum pooling layers, 2 fully connected layers, 1 softmax layer, and 1 output layer. Three different convolution kernels are used in the three convolutional layers: the first layer is to extract edge features, the second layer is to extract main contour features, and the third layer is to extract texture features; the input layer sets the size of the input layer according to the size of the injection molded part image, and the forward propagation of the calculation process from input data to output results; the fully connected layer refers to the use of the fully connected layer after multiple convolutional layers and pooling layers to map the learned feature representation to the label space of the sample, which is usually located in the last few layers of the network for classification or regression tasks; the output layer sets the number of neurons in the output layer according to the task requirements. For classification tasks, the number of neurons should be the same as the number of categories, and the softmax function is used to output the probability distribution.

[0047] S32, the forward transfer unit calculates the convolutional layer forward transfer formula "a'=σ( Wlm*Xm+bl), a' is the output of the lth layer, σ is the activation function, Wlm is the convolution kernel of the lth layer, Xm is the input feature map, bl is the bias term" to obtain the convolution layer output and pass it to the back propagation unit; Further explanation: the convolutional layer forward transfer calculation formula "a'=σ( Wlm*Xm+bl)" is calculated by "a²=σ(z²)=σ(a²*W²+b²)", where the superscript 2 is the number of layers 2, the asterisk * is convolution, b is bias, and σ is the activation function; in the convolutional neural network, the forward propagation includes: the data of the input layer is convolved through the convolution kernel to generate a feature map; the output of the convolution layer is nonlinearly transformed through the activation function; the output of the activation layer is downsampled through the pooling operation to reduce the number of parameters and the amount of calculation; the output of the pooling layer is flattened and input into the fully connected layer to finally get the output result.

[0048] S33, the back propagation unit obtains the output of the convolution layer and the fully connected layer according to the back propagation calculation formula "dz[l]=da[l]*g[l]′(z[l]), da[l] is the input data, da[l-1], dw[l], db[l] is the output data, g[l]′ is the activation function sigmoid derivative", and passes it to the linear transformation unit; Further explanation: g[l]′(z)=σ′(z)”; the output is dw[l]=dz[l]*a[l-1], db[l]=dz[l], da[l-1]=W[l]T*dz[l]; the back propagation of the convolution layer transfers the error term da[l]da[l] to the previous layer through the convolution operation, flips the convolution kernel and applies it to the error map, calculates the local gradient through the derivative of the activation function, and implements it through matrix operations. The input and the convolution kernel are deformed into matrices and matrix multiplication is performed, that is, the numbers in the convolution window are pulled into a row to form a column vector and matrix multiplication is performed; the back propagation of the fully connected layer calculates the gradient of the weight and the gradient of the bias through the chain rule. For each node, the error term is transferred to the node of the previous layer through the weight matrix, and the local gradient is calculated through the derivative of the activation function. The error term is multiplied by the output of the current layer and the input of the previous layer through matrix operations to update it in the direction of minimizing the loss function.

[0049] S34, the linear transformation unit obtains the linear transformation output according to the linear transformation calculation formula "y=W*X+b, X is the input data, y is the output data, W is the weight matrix, and b is the bias vector", and passes it to the ReLU function unit; It is further explained that the linear transformation calculation formula "y=W*X+b" is derived from the basic convolution operation formula "z(I,j)=(X*h)(I,j)=∑m∑nW(i+m,j+n)*h(m,n)", where X is the input image, h is the convolution kernel (filter), and z is the output feature map. The convolution kernel is slid on the input image, and the sum of the product of the local area at each position and the element of the convolution kernel is calculated. The weight matrix (W) and the bias vector (b) are realized by linearly combining the input data.

[0050] S35, the ReLU function unit obtains the output of the Tanh function according to the ReLU function calculation formula "f(x)=max(0,x), f(x) is the output of the ReLU function, and x is the input of the ReLU function", and passes it to the anti-overfitting unit; Further explanation: the input x of the ReLU function is the output of the convolution layer or other layers, which is a nonlinear transformation activation function; when the ReLU function is in forward propagation, the gradient is 1 when the input is greater than 0, otherwise it is 0. Since the calculation is simple, the problem of gradient disappearance will not occur, so it performs better in deep networks and maintains the stability of the gradient during the propagation process. For each element in the input x, if it is greater than 0, the element itself is output.

[0051] S36. The anti-overfitting unit obtains the TDropout output according to the Dropout calculation formula "y=f(W*X+b), y is the neural network output, f is the activation function, W is the weight matrix, X is the input data, and b is the bias vector)".

[0052] Further explanation: TDropout is a randomization technique used for neural network training or inference, which can effectively prevent overfitting. Overfitting occurs when a neural network is too complex to generalize to new data. It randomly disconnects some neurons on each layer of the neural network and keeps the connection of each neuron unchanged with a certain probability, thereby reducing the model's dependence on training data and thus reducing the risk of overfitting. It improves the generalization ability of the model on the test set and makes the model perform better on unseen data. It can make the neural network learn more robust features and still maintain good performance after abandoning some neurons.

[0053] S40, the model training module trains the model according to the collected data model and adjusts the model parameters to optimize the model performance to determine the process parameters, ensure the stability of the molding quality, and pass it to the injection molding module; See also Fig.11 As shown, the step S40 includes the following steps: S41, the normalization processing unit obtains the output of Batch Norm according to the Batch Norm calculation formula "h=Φ(BN(Wx+b)), h is the output of the fully connected layer, Φ is the activation function, BN is the batch normalization operator, W is the weight parameter, x is the input of the fully connected layer, and b is the bias parameter", and passes it to the convolution output unit; Further explanation: the “h = Φ (BN (Wx + b))” is obtained by “BN (x) = γ⊙ [(x - μ B ) / σ B ]+β”, where μ B is the mean, σ B is the standard deviation, γ is the stretch parameter, β is the offset parameter, (x-μ B ) / σ BTo standardize the normal distribution; the Batch Norm layer (referred to as "BN layer") is usually placed between the affine transformation and the activation function in the fully connected layer, and the output value of each layer in the network is standardized to make it more obedient to the normal distribution, which can accelerate the training speed of the neural network and help prevent gradient disappearance and gradient explosion; the BN layer reduces the correlation between input data by standardizing the data of each mini-batch, so that the mean of each sample is 0 and the variance is 1, thereby reducing the problem of internal covariate shift, helping to accelerate the training process, improve the stability and generalization ability of the model, and avoid reducing the problem of neural network representation by learning the scaling and offset parameters of each feature dimension.

[0054] S42, the convolution output unit calculates the convolution output according to the convolution output formula "O(n,k,ox,oy)= I(n,c,ix,iy)*W(k,c,r,s), O(n,k,ox,oy) convolution output data, C is the convolution channel, R is the convolution kernel height, S is the convolution kernel width, I(n,c,ix,iy), W(k,c,r,s) are respectively the convolution input data "to obtain the convolution output data, and pass it to the pooling calculation unit; It is further explained that the number of convolution channels C is obtained according to the design; the convolution kernel height R and the convolution kernel width S are calculated by the formulas "R=(R1-F+2P) / L+1" and "S=(S1-F+2P) / L+1", respectively, where R1 and R are the heights of the kernel before and after the convolution, F is the convolution kernel size, P is the image plus zero thickness, L is the distance of each movement, and S1 and S are the widths of the kernel before and after the convolution, respectively.

[0055] S43, the pooling calculation unit calculates the maximum pooling algorithm according to the formula "M(i,j)= A(i*s+u,j*s+v), M(i,j) is the (i,j)th element of the output feature map, u and v are indexes that vary in the range [0,f-1], and s is the step size. The spatial dimension of the feature map is obtained and passed to the gradient optimization unit; Further explanation: the maximum pooling algorithm refers to the use of overlapping 2*2 windows in the pooling layer to reduce the size of the feature map, thereby reducing the amount of calculation, while retaining the most significant features, and reducing the spatial dimension (i.e., width and height) of the feature map by taking the maximum value in the local area, but without changing the number of channels, and selecting the largest value in a local window of the input feature map as the output of the window. Its parameters include: the size of the pooling kernel, step size, padding, expansion, and whether to round up. The spatial dimension of the feature map can be reduced while reducing the size of the feature map, reducing the amount of calculation and the risk of overfitting; by taking the maximum value in the local area, the most important significant feature information is retained; translation invariance and scale invariance have a certain degree of robustness to slight changes in the input; the step size s is the distance that the pooling window moves on the input feature map; the pooling types include maximum pooling or average pooling. After the convolution layer, the pooling layer is used to reduce the dimension of the feature map, reduce the amount of calculation, and prevent overfitting, including maximum pooling and average pooling.

[0056] S44, the gradient optimization unit obtains the gradient descent data according to the gradient descent method calculation formula "Xn+1=Xn-α*▽f(Xn), Xn is the nth iteration parameter value, α is the learning rate or step size, ▽f(Xn) is the gradient of the function f at Xn", and passes it to the training learning unit; It is further explained that the gradient descent method can gradually approach the optimal solution by iteratively updating parameters, so that the minimized differentiable function f(x) can obtain the minimum value x. In the formula, the learning rate or step size α is used to control the step size of each iteration, and the gradient ▽f(Xn) of the function f at Xn, that is, the derivative, starts from any initial point, and determines the next position according to the gradient direction of the current position (that is, the direction in which the function decreases fastest) and the learning rate, and iterates continuously until the stopping condition is met, including: starting from any initial point, calculating the gradient ▽f(Xn) of the current point, updating the parameter Xn+1=Xn-α*▽f(Xn), checking the stopping condition, and ending the iteration if it is met, otherwise recalculating the gradient.

[0057] S45, the training learning unit determines the training model according to the gradient descent data, and divides it into a training set, a validation set, and a test set, performs model training, and passes it to the test evaluation unit; It is further explained that the loss function is minimized by continuously adjusting the model parameters to improve the prediction accuracy of the model to ensure that the model can accurately identify defects of injection molded parts; wherein, the training set refers to a data set that can be used to train the model, fit the data analysis rules, that is, determine the various learning parameters such as the model weight and bias, and can be used multiple times; the validation set refers to a data set that can verify the performance of the model. During the model training process, the model parameters and hyperparameters can be adjusted to optimize the model performance and avoid overfitting, and the model is selected, but it does not participate in the determination of the learning parameters, and selects model parameters and hyperparameters with smaller model errors, and can be used multiple times; the test set refers to a data set that can evaluate the performance of the model. After the model training is completed, the generalization ability of the model on unknown data is evaluated. It is used once after training to evaluate the effect of the final model. It does not participate in the learning parameter process or the hyperparameter selection process.

[0058] S46. The test evaluation unit obtains the F1 comprehensive evaluation score according to the training model comprehensive evaluation score calculation formula "F1=2*(Ac*Re) / (Ac+Re), Ac=(TP+TN) / (TP+FP+FN+TN), Re=TP / (TP+FN), F1 is the training model evaluation score, Ac is the precision, Re is the recall, TP is the number of samples that are actually positive and correctly predicted to be positive, FP is the number of samples that are actually negative but incorrectly predicted to be positive, FN is the number of samples that are actually positive but incorrectly predicted to be negative, TN is the number of samples that are actually negative and correctly predicted to be negative" and transmits it to the processing center; It is further explained that after the model training is completed, it is necessary to test it to verify the accuracy and reliability of the model. Problems with the model may be found during the test, and corresponding optimization and adjustment are required; the training model evaluation score, namely the F1 value, is the harmonic mean of precision and recall, which can evaluate the performance of the classification model. The value range is 0 to 1, where 1 is the best performance and 0 is the worst performance. The higher the F1 value, the better the prediction effect of the model, and vice versa. Therefore, if the F1 value is close to 1, it indicates that the model performs well while maintaining a balance between precision and recall; the precision is the prediction of the classification model. The ratio of the true proportion in the result measures the proportion of samples predicted to be positive that are actually positive examples, and its value range is from 0 to 1. The closer its value is to 1, the stronger the model's ability to predict positive examples is, and the closer its value is to 0, the weaker the model's ability to predict positive examples is; the recall rate measures the proportion of positive examples that the model can correctly identify to all actual positive examples, and its value range is from 0 to 1. The closer its value is to 1, the lower the false negative rate of the model's predicted positive examples is, and the closer its value is to 0, the higher the false negative rate of the model's predicted positive examples is; adjust the hyperparameters according to the model's learning rate, regularization coefficient and other performance to optimize the model performance.

[0059] S47. The processing center compares the actual evaluation score of the model test with the evaluation score standard of the injection molded part abnormal recognition training model test stored in the memory: if it meets the standard, it is determined as a molding process parameter to notify the injection molding production; if it does not meet the standard, it is transmitted to the alarm and notified to continue training until it meets the standard.

[0060] It is further explained that the best standard for the injection molded parts abnormality recognition training model test evaluation score is greater than 0.8, which is determined according to the molding performance of the company's plastic materials, the cost of plastic materials and the customer's requirements for the appearance quality of injection molded parts. Some customers allow structural injection molded parts to have some defects in appearance, but have very strict requirements on size or deformation. Therefore, the standard for the injection molded parts abnormality recognition training model evaluation score in this case is lower; some customers require that those appearance-related injection molded parts must not have any other lines or dots on the appearance and must be the same color. In this case, the standard for the injection molded parts abnormality recognition training model evaluation score is 1 or close to 1.

[0061] S50, the injection molding module performs injection molding according to the process parameters predicted by the trained model and obtains various quality information of the injection molded parts, and transmits it to the processing center; See also Fig.12 As shown, the step S50 includes the following steps: S51, the molding application unit automatically inputs the system according to the determined molding process parameters and debugs them in place for injection molding production, ensuring that the quality information of the actual injection molded parts is consistent with the model, and transmits it to the appearance inspection unit; Further explanation: the trained model is applied to injection molding production, including the integration of predetermined injection molding process parameters into the existing production control system to support the operation of the prediction model. According to the injection molding instructions of the target injection molded parts, the corresponding molding process parameter data is matched to the system, and the injection molding machine is controlled to perform injection molding according to the set molding process parameters. However, due to abnormalities in the unstable material or abnormalities in the machine during the process, it cannot be in place at one time. The process parameters are optimized in time through real-time prediction of the model, thereby improving production efficiency and product quality. When the quality information of the injection molded parts is inconsistent with the predetermined injection molded parts quality standards, the system automatically adjusts the process parameters to control the injection molding machine to re-inject and mold until it is consistent with the predetermined injection molded parts quality standards, but it does not affect the injection molding time. After the quality abnormality is found, the system will automatically adjust the machine; if there is no abnormality in the material, machine and its supporting equipment, the process parameters after the injection molded parts are qualified are completely consistent with the trained model.

[0062] S52, the appearance inspection unit obtains the appearance image of the molded injection molded part through a set high-definition camera or image sensor and converts it into recognizable various quality information of the injection molded part appearance, and transmits it to the performance inspection unit; It is further explained that the high-definition camera can take 360-degree high-definition photos from different angles and directions to obtain the appearance image of the injection molded part, or it can use an image sensor to collect, and perform pre-processing such as removing noise, adjusting contrast, and performing color correction on the obtained injection molded part image to improve the image quality of the injection molded part and highlight the features of the injection molded part, so as to ensure that the injection molded part image is clear and visible, which is convenient for subsequent feature extraction; the key feature information such as color, texture, shape, etc. of the injection molded part image after pre-processing is extracted, and these features are quantified through image algorithms to facilitate computer Identification and processing; input the extracted feature information into the above-mentioned trained model for classification and identification of the appearance features of injection molded parts, and classify them into different categories such as no defects and multiple defects such as insufficient filling, overflow, pores, ripples / silver streaks, surface halos, fusion lines / bubbles, etc.; and select injection molded parts with one or more defects such as insufficient filling, overflow, pores, ripples / silver streaks, surface halos, fusion lines / bubbles, etc. from qualified products, and read the appearance of the injection molded parts for classification and marking, etc. Injection molded parts without any defects are qualified injection molded parts and are specially marked.

[0063] S53. The performance testing unit obtains the dimension data, physical performance test data, and chemical performance test data of the injection molded parts with normal appearance through the injection molded parts quality inspection equipment to meet the customer's needs for injection molded parts.

[0064] Further explanation, the injection molded parts quality inspection equipment includes: dimensional deviation testing equipment, physical property testing equipment, chemical property testing equipment; through image measuring instruments, three-coordinate measuring instruments and other dimensional deviation testing equipment to obtain various dimensional data such as size, roundness, straightness, verticality, parallelism, angle, etc. of injection molded parts; through melt flow rate meter, tensile testing machine, impact testing machine, thermal analyzer, infrared spectrometer and gas chromatograph and other physical property testing equipment to obtain injection molded parts of fluidity (density, hardness, etc.), mechanical properties (tensile strength, bending strength, compression strength, yield strength, impact strength, toughness, wear resistance, etc.), thermal properties (thermogravimetry, thermal conductivity, high temperature resistance, cold resistance, etc.) and other physical property test data; through infrared spectrometer, X-ray diffractometer, thermogravimetric analyzer, oxygen index meter and other chemical property testing equipment to obtain injection molded parts of chemical composition (corrosion resistance, etc.), structure / phase, combustion performance, acid and alkali resistance, pollution resistance, water vapor resistance, mildew resistance and other chemical property test data.

[0065] S60, the processing center compares the quality information of the actual injection molded parts with the quality standards of the injection molded parts stored in the memory: if the standards are met, production continues; if the standards are not met, the information is transmitted to the alarm and a notice is given to rework.

[0066] It is further explained that the quality standards of injection molded parts include no abnormal appearance, no dimensional deviation, physical properties meeting the standards, and chemical properties meeting the standards, among which the test data of physical properties and chemical properties are based on the incoming material procurement inspection data of plastic raw materials and the sampling data of injection molded parts; the dimensional deviation of injection molded parts is based on the sampling data, and the various quality information of appearance is obtained through real-time detection during production, and is stored in the memory through wireless or wired networking of various injection molded parts quality inspection equipment for the system to call at any time; when quality abnormalities occur during the production process due to abnormal plastic materials, or machine abnormalities, or mold wear, or human intervention, the system will automatically alarm and notify the system to repair or improve; in this way, feedback data on the quality of injection molded parts is collected, and the model is continuously optimized and improved to improve the accuracy and efficiency of screening, and the trained and maturely applied model is converted into a specific format, and the code for running these processes is written according to the injection molding supporting equipment, thereby forming a complete artificial intelligence model training process to ensure the accuracy, practicality and safety of the model.

[0067] It is further explained that the above steps are displayed in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows, unless clearly stated in this document. There is no strict order restriction for the execution of these steps, and these steps can also be executed according to other orders, and some steps can include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed and completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0068] It is further explained that the present invention is described according to the content implemented by the software program of the zero-defect quality control system for injection molded parts production. Each zero-defect quality control method for injection molded parts production is divided into several modules or units to implement the software program instructions generated by each step. The software program instructions include the above-mentioned zero-defect quality control method for injection molded parts production.

[0069] A zero-defect quality control system for injection molded parts production provided by the present invention also includes a computer-readable storage medium containing a memory; the memory stores a computer program, and when the above functional modules execute the computer program, the steps of the zero-defect quality control method for injection molded parts production described in any one of the above are implemented; the computer-readable storage medium stores a computer program, and when the computer program is executed by each functional module, the steps of the zero-defect quality control method for injection molded parts production described in any one of the above are implemented.

[0070] It is further explained that the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two, and may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, apparatus or device, including but not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above, wherein the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, carrying a computer-readable program code, and the propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination of the above, and may also be any computer-readable medium other than a computer-readable storage medium, and the computer-readable signal medium may send, propagate or transmit a program for use by or in combination with an instruction execution system, apparatus or device, and the program code that may be contained may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0071] It is further explained that the computer program is divided into multiple modules / units, and the multiple modules / units are stored in the memory and executed by each module / unit to complete the present invention; the multiple modules / units can be a series of computer program instructions that can complete specific functions, and the instructions are used to describe the execution process of the computer program in each module / unit.

[0072] It is further explained that some of the contents described above in the present invention are written in accordance with the contents implemented by the software copyrights such as "Manufacturing Resource Digital Information Coordination Management System (Registration Number: 2016SR093182)", "Electronic Manufacturing Kanban Management System (Registration Number: 2016SR118737)", "Communication Terminal Production Site Process Control System (Registration Number: 2016SR118741)", and "Pace-based Zero Defect Production Control System (Registration Number: 2019SR0131813)" applied by our company on May 4, 2016.

[0073] The present invention also provides a zero-defect quality control device for injection molded parts production, which is implemented by the above-mentioned zero-defect quality control method for injection molded parts production.

[0074] It is further explained that the control device is composed of the above-mentioned related equipment or devices, and can be made into a complete set of zero-defect production line for injection molded parts together with injection molding production and detection equipment according to the production needs of the injection molding enterprise; it can also be made into a unit device of each functional module belonging to each step or the entire control device, and each control device can be connected to the injection molding production and detection equipment through a wireless connection during installation; the existing injection molding production and detection equipment can also be upgraded and modified by adding the control device, thereby saving equipment costs. These equipment structures are not described in detail here.

[0075] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is defined by the attached claims rather than the above description, and therefore it is intended that all changes that fall within the meaning and scope of the equivalent elements of the claims are included in the present application.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application can be modified or replaced by equivalents, and all of these should be included in the scope of protection of the present application.

Claims

1. A method for zero-defect quality control of injection molded parts production, characterized in that: The method is applied to a zero-defect quality control system for injection molded parts production, the system comprising an information acquisition module, a data processing module, a model building module, a model training module, an injection molding module, a wireless communication module, a memory, an alarm, a processing center, and an intelligent mobile terminal; the information acquisition module, the data processing module, the model building module, the model training module, the injection molding module, the wireless communication module, the memory, and the alarm are respectively connected to the processing center; the intelligent mobile terminal comprises a smart phone, a tablet computer or an intelligent remote control, which are respectively connected to the wireless network of the wireless communication module within the range of a wireless network or the Internet; The method comprises the following steps: S10. During production, the information collection module collects various characteristic information and historical molding process data of the injection mold to be formed and processes them for subsequent model construction, and transmits them to the data processing module; S20, the data processing module standardizes, smoothes, filters, removes noise, and classifies the mined data for subsequent model construction, and passes it to the model construction module; S30, the model building module builds an injection molding abnormality type prediction model and its required training set and test set according to the identified abnormal pattern through a convolutional neural network algorithm, and passes it to the model training module; S40, the model training module trains the model according to the collected data model and adjusts the model parameters to optimize the model performance to determine the process parameters, ensure the stability of the molding quality, and pass it to the injection molding module; S50, the injection molding module performs injection molding according to the process parameters predicted by the trained model and obtains various quality information of the injection molded parts, and transmits it to the processing center, including the following steps: S51, the molding application unit automatically inputs the system according to the determined molding process parameters and debugs them in place for injection molding production, ensuring that the quality information of the actual injection molded parts is consistent with the model, and transmits it to the appearance inspection unit; S52, the appearance inspection unit obtains the appearance image of the molded injection molded part through a set high-definition camera or image sensor and converts it into recognizable various quality information of the injection molded part appearance, and transmits it to the performance inspection unit; S53. The performance testing unit obtains the dimension data, physical performance test data, and chemical performance test data of the injection molded parts with normal appearance through the injection molded parts quality testing equipment to meet the customer's demand for injection molded parts; S60, the processing center compares the quality information of the actual injection molded parts with the quality standards of the injection molded parts stored in the memory: if the standards are met, production continues; if the standards are not met, the information is transmitted to the alarm and a notice is given to rework.

2. A method for zero-defect quality control of injection molded parts production as claimed in claim 1, characterized in that: The system also includes: the wireless communication module is provided with a wireless network unit, which is responsible for sending and receiving wireless signals, and is automatically networked with the intelligent control terminal within the effective network range; the alarm compares the actual evaluation score of the model test with the injection molded part abnormality identification training model test evaluation score standard stored in the memory, and automatically issues a sound alarm and notifies to continue training if the standard is not met; and compares the actual various quality information of the injection molded parts with the injection molded part quality standard stored in the memory, and automatically issues a sound alarm and notifies to rework if the standard is not met; the memory is responsible for the information collection module, the data processing module, the model construction module, the model training module, the injection molding module, the wireless communication module, the alarm information storage, and the injection molded part abnormality identification training model test evaluation score standard and the injection molded part quality standard storage; The processing center is responsible for information collection module, data processing module, model building module, model training module, injection molding module, wireless communication module, alarm, and information transmission of memory. It is the hub center of the system and compares the actual evaluation score of the model test with the injection molding abnormality recognition training model test evaluation score standard stored in the memory: if it meets the standard, it is determined as a molding process parameter to notify the injection molding production; if it does not meet the standard, it is transmitted to the alarm and notified to continue training; and compares the various quality information of the actual injection molding parts with the quality standard of the injection molding parts stored in the memory: if it meets the standard, production continues; if it does not meet the standard, it is transmitted to the alarm and notified to rework; The information collection module includes a mold information unit, a molding information unit, and a data mining unit, which is responsible for collecting various characteristic information and historical molding process data of the injection mold to be molded and processing them for subsequent model construction, and passing them to the data processing module; The data processing module includes a standard data unit, a smoothing processing unit, a filtering and denoising unit, and a data analysis unit, which are responsible for standardizing, smoothing, filtering and denoising, and classifying and identifying the mined data for subsequent model construction and passing it to the model construction module; The model building module includes a network structure unit, a forward transmission unit, a back-propagation unit, a linear transformation unit, a ReLU function unit, and an anti-overfitting unit, which is responsible for building an injection molding abnormality type prediction model and its required training set and test set through a convolutional neural network algorithm according to the identified abnormal pattern, and passing them to the model training module; The model training module includes a normalization processing unit, a convolution output unit, a pooling calculation unit, a gradient optimization unit, a training learning unit, and a test evaluation unit. The model is trained according to the collected data model and the model parameters are adjusted to optimize the model performance to determine the process parameters, ensure the stability of the molding quality, and pass them to the injection molding module; The injection molding module includes a molding application unit, an appearance inspection unit, and a performance testing unit. It performs injection molding according to the process parameters predicted by the trained model and obtains various quality information of the injection molded parts, and transmits it to the processing center.

3. A method for zero-defect quality control of injection molded parts production as claimed in claim 1, characterized in that: The step S10 comprises the following steps: S11, the mold information unit obtains characteristic information of various injection molds to be molded through the intelligent sensor and data collector set on the injection molding machine, so as to collect subsequent molding parameters and transmit them to the molding information unit; S12, the molding information unit obtains various molding data of the corresponding injection mold by networking and matching with the industry database according to the collected mold feature information, so as to facilitate subsequent data processing and transmit it to the data mining unit; S13, the data mining unit calculates the formula "y'=sign(wd*xd+wd-1*xd-1), E(w)=1 / 2* (yi-y')², y' is the predicted output, w is the perceptron model parameter, x is the linear perceptron model attribute, and yi is the actual output" to obtain the cleaned and mined data.

4. A method for zero-defect quality control of injection molded parts production as claimed in claim 1, characterized in that: The step S20 comprises the following steps: S21, the standard data unit integrates, removes duplication, removes differences, unifies, and pre-processes the mined data to ensure data standardization for subsequent smoothing processing, and transmits it to the smoothing processing unit; S22, the smoothing processing unit obtains the smoothed data after mining according to the exponential smoothing calculation formula "St=α*yt-1+(1-α)St-1, St is the smoothing value of the tth period, yt-1 is the actual value of the t-1th period, and α is the smoothing coefficient" to ensure the integrity of the data for subsequent filtering and denoising, and pass it to the filtering and denoising unit; S23, the filtering and denoising unit obtains the filtered and denoised shaped data according to the first-order low-pass filtering calculation formula "y(t)=y(t-1)+K*[u(t)-y(t-1)], y(t) is the output signal at the current moment, K is the signal balance coefficient (0-1), u(t) is the input signal at the current moment, and y(t-1) is the output signal at the previous moment", and transmits it to the data analysis unit; S24. The data analysis unit uses deep learning algorithms to conduct in-depth analysis of the collected data and identify six potential abnormal patterns of injection molded parts, optimizing the production process for subsequent model construction.

5. A method for zero-defect quality control of injection molded parts production as claimed in claim 1, characterized in that: The step S30 comprises the following steps: S31, the network structure unit identifies the abnormal pattern classification label of the injection molded part and establishes the network structure of the input layer, convolution layer, pooling layer, full connection layer, and output layer, and transmits it to the forward transmission unit; S32, the forward transfer unit calculates the forward transfer formula of the convolution layer according to "a'=σ( Wlm*Xm+bl), a' is the output of the lth layer, σ is the activation function, Wlm is the convolution kernel of the lth layer, Xm is the input feature map, bl is the bias term" to obtain the convolution layer output and pass it to the back propagation unit; S33, the back propagation unit obtains the output of the convolution layer and the fully connected layer according to the back propagation calculation formula "dz[l]=da[l]*g[l]′(z[l]), da[l] is the input data, da[l-1], dw[l], db[l] is the output data, g[l]′ is the activation function sigmoid derivative", and passes it to the linear transformation unit; S34, the linear transformation unit obtains the linear transformation output according to the linear transformation calculation formula "y=W*X+b, X is the input data, y is the output data, W is the weight matrix, and b is the bias vector", and passes it to the ReLU function unit; S35, the ReLU function unit obtains the output of the Tanh function according to the ReLU function calculation formula "f(x)=max(0,x), f(x) is the output of the ReLU function, and x is the input of the ReLU function", and passes it to the anti-overfitting unit; S36. The anti-overfitting unit obtains the TDropout output according to the Dropout calculation formula "y=f(W*X+b), y is the neural network output, f is the activation function, W is the weight matrix, X is the input data, and b is the bias vector)".

6. A method for zero-defect quality control of injection molded parts production as claimed in claim 1, characterized in that: The step S40 comprises the following steps: S41, the normalization processing unit calculates the Batch Norm formula "h = Φ (BN (Wx + b)), where h is the output of the fully connected layer, Φ is the activation function, BN is the batch normalization operator, W is the weight parameter, x is the input of the fully connected layer, and b is the bias parameter. The output of Batch Norm is obtained and passed to the convolution output unit; S42, the convolution output unit calculates the convolution output according to the formula "O(n,k,ox,oy)= I(n,c,ix,iy)*W(k,c,r,s), O(n,k,ox,oy) convolution output data, C is the convolution channel, R is the convolution kernel height, S is the convolution kernel width, I(n,c,ix,iy), W(k,c,r,s) is the convolution input data" to obtain the convolution output data and pass it to the pooling calculation unit; S43, the pooling calculation unit calculates the maximum pooling algorithm according to the formula "M(i,j)= A(i*s+u,j*s+v), M(i,j) is the (i,j)th element of the output feature map, u and v are indexes that vary in the range [0,f-1], and s is the step size. The spatial dimension of the feature map is obtained and passed to the gradient optimization unit; S44, the gradient optimization unit obtains the gradient descent data according to the gradient descent method calculation formula "Xn+1=Xn-α*▽f(Xn), Xn is the nth iteration parameter value, α is the learning rate or step size, ▽f(Xn) is the gradient of the function f at Xn", and passes it to the training learning unit; S45, the training learning unit determines the training model according to the gradient descent data, and divides it into a training set, a validation set, and a test set, performs model training, and passes it to the test evaluation unit; S46, the test evaluation unit calculates the comprehensive evaluation score of the training model according to the formula "F1=2*(Ac*Re) / (Ac+Re),Ac= (TP+TN) / (TP+FP+FN+TN), Re=TP / (TP+FN), F1 is the training model evaluation score, Ac is the precision rate, Re is the recall rate, TP is the number of samples that are actually positive and correctly predicted as positive, FP is the number of samples that are actually negative but wrongly predicted as positive, FN is the number of samples that are actually positive but wrongly predicted as negative, TN is the number of samples that are actually negative and correctly predicted as negative" to obtain the F1 comprehensive evaluation score and pass it to the processing center; S47. The processing center compares the actual evaluation score of the model test with the evaluation score standard of the injection molded part abnormal recognition training model test stored in the memory: if it meets the standard, it is determined as a molding process parameter to notify the injection molding production; if it does not meet the standard, it is transmitted to the alarm and notified to continue training until it meets the standard.

7. A method for zero-defect quality control of injection molded parts production as claimed in any one of claims 1 to 6, characterized in that: The system also includes a computer-readable storage medium containing a memory; the memory stores a computer program, and when the above functional modules execute the computer program, the steps of the zero-defect quality control method for injection molded parts production described in any one of claims 1 to 6 are implemented; the computer-readable storage medium stores a computer program, and when the computer program is executed by the functional modules, the steps of the zero-defect quality control method for injection molded parts production described in any one of claims 1 to 6 are implemented.

8. A zero-defect quality control device for injection molded parts production, implemented by a zero-defect quality control method for injection molded parts production as described in any one of claims 1 to 6 above.

Images