Environment-friendly high-glowing filament high-impact PBT (polybutylene terephthalate) particle and preparation method thereof

Abstract

The invention relates to the field of PBT particle preparation, in particular to an environment-friendly high-glowing filament high-impact PBT particle and a preparation method thereof.The method comprises the steps that the actual feeding speed of flame-retardant powder, the actual feeding speed of glass fibers and the discharging speed of equipment are constructed through a multi-scale CNN and an attention mechanism to generate feeding feature vectors; performing feeding fluctuation rule capture on the feeding feature vector through an LSTM model to generate feeding bridging caking features; the feed bridging caking features are passed through a fully connected network to generate a predicted powder feed speed and a predicted fiber feed speed, and a desired powder feed speed and a desired fiber feed speed of the forced feeding equipment are adjusted. According to the method, the feeding speed capable of actively predicting feeding fluctuation and inhibiting bridging and caking is generated through a multi-model collaborative architecture and mapping, so that the dispersion uniformity of feeding components is ensured, and the stable preparation performance of the environment-friendly high-glow-wire high-impact PBT particles is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of PBT particle preparation, in particular to an environmentally friendly high-glow wire high-impact PBT particle and a preparation method thereof. BACKGROUND

[0002] Polybutylene terephthalate (PBT) is a high-performance engineering plastic with excellent mechanical properties, heat resistance and molding processability, and is widely used in electronic appliances, automobile manufacturing and other fields. With the increasing demand for material safety and reliability in downstream industries, high-glow wire flame retardant and impact resistance of PBT particles are required.

[0003] For environmentally friendly high-glow wire high-impact PBT particles that do not contain harmful ingredient triphenyl phosphate (TPP), they mostly contain composite flame retardants and glass fibers. The composite flame retardant system achieves efficient flame retardation, and the glass fiber can improve the material rigidity and dimensional stability, and improve the impact performance.

[0004] However, in the process of adding the composite flame retardant and the glass fiber to the PBT resin to make the PBT particles, the feeding speed setting accuracy of the forced feeding equipment determines the uniformity of the components, which becomes a key bottleneck affecting the final performance of the PBT particles. Specifically, the composite flame retardant is mostly in the form of powder, and the glass fiber is in the form of short cut filament. Due to the material characteristics of powder agglomeration due to moisture absorption, fiber entanglement and equipment parameter fluctuation, at a fixed feeding speed, bridging (blockage of the feeding channel) or caking (powder agglomerates) problems occur, which reduces the uniformity of the components in the feeding mixture and reduces the flame retardant and impact resistance of the PBT particles.

[0005] Currently, the feeding control of the forced feeding equipment mostly relies on the real-time feedback adjustment of the loss-on-ignition balance. The actual feeding rate is compared with the target rate to adjust the feeding speed of the equipment. However, this process is a passive response, which cannot predict sudden caking of the powder and other short-term fluctuation trends, and is difficult to capture the potential impact of changes in material characteristics on feeding, resulting in a long lag time for feeding speed adjustment. Especially in the high-capacity continuous production of the forced feeding equipment, it is easy to cause large batch differences in the performance of PBT particle products, which further affects the product qualification rate in downstream high-performance application scenarios.

[0006] Therefore, how to actively predict the feeding fluctuation of the composite flame retardant and the glass fiber, suppress the bridging and caking of the feeding speed adjustment in the preparation process of the environmentally friendly high-glow wire high-impact PBT particle, and then ensure the uniformity of the components, realize the stable performance of the product preparation of the PBT particle is a technical problem to be solved at present. SUMMARY

[0007] To this end, the present application provides an environmentally friendly high glowing wire high impact PBT particle and a preparation method thereof. Through the multi-model collaborative architecture of CNN, attention mechanism and LSTM, the short-term key features of the actual feeding speed of the composite flame retardant and glass fiber during the preparation of the PBT particle are realized, the characteristic signals of the bridging and caking are identified, and then the feeding speed capable of actively predicting the feeding fluctuation and inhibiting the bridging and caking is generated, thereby ensuring the uniformity of component dispersion and realizing the stable preparation performance of the environmentally friendly high glowing wire high impact PBT particle.

[0008] To achieve the above-mentioned purpose, the present application provides a preparation method of an environmentally friendly high glowing wire high impact PBT particle, comprising:

[0009] The composite flame retardant powder and glass fiber are added to the forced feeding equipment, and after the PBT resin is forcedly fed, PBT particle premix is obtained, and the PBT particle premix is extruded and injection molded to obtain PBT particles;

[0010] The forced feeding equipment obtains the actual feeding speed of the flame retardant powder, the actual feeding speed of the glass fiber and the equipment discharge speed through multiple loss weight scales;

[0011] The actual feeding speed of the flame retardant powder, the actual feeding speed of the glass fiber and the equipment discharge speed are subjected to short-term feature extraction by a feature extraction model based on a multi-scale CNN and an attention mechanism architecture to generate a feeding feature vector;

[0012] The feeding feature vector is subjected to feeding law capture by an LSTM model to generate a feeding bridging and caking feature;

[0013] The feeding bridging and caking feature is subjected to full connection network to generate a predicted powder feeding speed and a predicted fiber feeding speed, and the expected powder feeding speed and the expected fiber feeding speed of the forced feeding equipment are adjusted based on the predicted powder feeding speed and the predicted fiber feeding speed.

[0014] Further, the feature extraction model includes a powder feature extraction branch and a fiber feature extraction branch, the feeding feature vector includes a powder feeding feature vector and a fiber feeding feature vector, and the process of generating the feeding feature vector by short-term feature extraction by the feature extraction model includes:

[0015] The actual feeding speed of the flame retardant powder, the powder target feeding speed and the actual running data of the feeding machine are taken as powder branch input data, and the powder branch input data is subjected to powder multi-scale convolution layer of the powder feature extraction branch to generate multi-scale powder feeding features;

[0016] The multi-scale powder feeding characteristics and the powder branch input data are guided and weighted by the powder attention mechanism of the powder feature extraction branch to generate the powder feeding feature vector;

[0017] The glass fiber actual feeding speed, fiber target feeding speed and feeder actual operation data are taken as the fiber branch input data, and the fiber branch input data is guided and weighted by the fiber multi-scale convolution layer of the fiber feature extraction branch to generate multi-scale fiber feeding characteristics;

[0018] The multi-scale fiber feeding characteristics and fiber branch input data are guided and weighted by the fiber attention mechanism of the fiber feature extraction branch to generate the fiber feeding feature vector.

[0019] Further, the multi-scale powder feeding characteristics include powder fluctuation characteristics, powder bridging trend characteristics and powder flow characteristics, and the process of generating the powder feeding feature vector by the powder multi-scale convolution layer and the powder attention mechanism includes:

[0020] The powder branch input data is sequentially guided and weighted by the first powder convolution layer, the second powder convolution layer and the third powder convolution layer included in the powder multi-scale convolution layer to generate the powder fluctuation characteristics, the powder bridging trend characteristics and the powder flow characteristics;

[0021] The standard deviation, skewness and kurtosis of the powder branch input data are respectively guided and weighted by the powder attention mechanism to generate fluctuation attention weight, caking attention weight and bridging attention weight, and the fluctuation attention weight, caking attention weight and bridging attention weight are weighted and summed to generate the powder feeding feature vector;

[0022] The multi-scale fiber feeding characteristics include fiber breakage characteristics, fiber winding trend characteristics and fiber conveying characteristics, and the process of generating the fiber feeding feature vector by the fiber multi-scale convolution layer and the fiber attention mechanism includes:

[0023] The fiber branch input data is sequentially guided and weighted by the first fiber convolution layer, the second fiber convolution layer and the third fiber convolution layer of the fiber multi-scale convolution layer to generate the fiber breakage characteristics, the fiber winding trend characteristics and the fiber conveying characteristics;

[0024] The mean, variance and information entropy of the fiber branch input data are respectively input into a fiber attention mechanism to generate a fracture attention weight, a winding attention weight and a feeding attention weight, and the fracture attention weight, the winding attention weight and the feeding attention weight are weighted and summed on the fiber fracture feature, the fiber winding trend feature and the fiber feeding feature to generate the fiber feeding feature vector;

[0025] The size of the convolution kernel of the convolution layer is sequentially from small to large: the first powder convolution layer, the first fiber convolution layer, the second powder convolution layer, the second fiber convolution layer, the third powder convolution layer and the third fiber convolution layer.

[0026] Further, the LSTM model includes a powder fast LSTM branch, a fiber slow LSTM branch and an interaction layer, and the process of generating the bridging and caking feature of the feeding frame by the LSTM model includes:

[0027] The powder feeding feature vector and its change amount are respectively input into the powder fast LSTM branch with a fiber interaction input gate and a change rate enhanced forgetting gate to generate a powder hidden state.

[0028] The fiber feeding feature vector and its change amount are respectively input into the fiber slow LSTM branch with a trend enhanced input gate and a smoothing forgetting gate to generate a fiber hidden state.

[0029] The powder hidden state and the fiber hidden state are input into the interaction layer to generate the bridging and caking feature of the feeding frame.

[0030] Further, the process of generating the powder hidden state includes:

[0031] The concatenation vector of the powder hidden state at the last time step, the powder feeding feature vector and the fiber hidden state at the last time step is linearly mapped to generate the fiber interaction input gate and the initial forgetting gate, the difference value of the powder feeding feature vector at multiple time steps is weighted and summed on the initial forgetting gate to generate the change rate enhanced forgetting gate, and the powder hidden state at the current time step is updated based on the fiber interaction input gate and the change rate enhanced forgetting gate.

[0032] The process of generating the fiber hidden state includes:

[0033] A glass fiber feeding trend value is calculated based on a difference between the glass fiber actual feeding speed and the fiber target feeding speed in a set time window, an initial input gate based on the fiber feeding feature vector and the glass fiber feeding trend value are weighted and summed to generate a trend enhanced input gate, a fiber initial forgetting gate based on the fiber feeding feature vector and a change rate of the fiber feeding feature vector are weighted and added to generate a smoothing forgetting gate, and the fiber hidden state of the current time step is updated based on the trend enhanced input gate and the smoothing forgetting gate.

[0034] Further, a process of generating the feeding bridge caking feature through the interaction layer includes:

[0035] The powder hidden state and the fiber hidden state are weighted and summed to generate a linear term;

[0036] The powder hidden state and the fiber hidden state are multiplied element by element to obtain an interaction term;

[0037] The linear term and the interaction term are weighted and added to generate a comprehensive hidden state, and the comprehensive hidden state, the powder hidden state and the fiber hidden state of the final time step are taken as the feeding bridge caking feature.

[0038] Further, a process of adjusting the expected powder feeding rotation speed and the expected fiber feeding rotation speed includes:

[0039] A powder adjustment amount is generated based on a difference between the predicted powder feeding speed and a powder target feeding speed, and the expected powder feeding rotation speed is calculated based on the powder adjustment amount and the current powder feeding rotation speed;

[0040] A fiber adjustment amount is generated based on a difference between the predicted fiber feeding speed and a fiber target feeding speed, and the expected fiber feeding rotation speed is calculated based on the fiber adjustment amount and the current fiber feeding rotation speed.

[0041] Further, the preparation method further includes:

[0042] A main loss term is constructed based on a mean square error between the predicted powder feeding speed and a sample true powder optimal feeding speed, and a mean square error between the predicted fiber feeding speed and a sample true fiber optimal feeding speed;

[0043] A flow stability loss term is constructed based on a ratio of the predicted powder feeding speed and the predicted fiber feeding speed, a ratio of a powder theoretical flow and a fiber theoretical flow;

[0044] A comprehensive loss function is constructed based on a weighted sum of the main loss term and the flow stability loss term, and the feature extraction model, the LSTM model and the fully connected network are collaboratively optimized and trained through the comprehensive loss function.

[0045] Especially, the intelligent feeding control based on the multi-scale CNN, attention mechanism, LSTM and fully connected network solves the proportion imbalance problem caused by powder bridging and fiber winding in traditional feeding.

[0046] The application also provides an environmentally friendly PBT particle with high glowing wire and high impact, which is prepared by the preparation method and comprises, by weight fraction, 30-50 parts of PBT resin, 5-40 parts of glass fiber and 23.9-60.4 parts of composite flame retardant.

[0047] The halogen-based flame retardant comprises, by weight fraction, 8-15 parts of brominated epoxy, 5-15 parts of brominated polystyrene and 1-5 parts of antimony trioxide.

[0048] The nitrogen-based MCA flame retardant comprises, by weight fraction, 3-10 parts of nitrogen-based flame retardant MCA, 2-6 parts of stannate and 1-3 parts of silicate.

[0049] Further, the auxiliary additive comprises, by weight fraction, 3-5 parts of toughening agent, 0.2-0.4 parts of antioxidant, 0.5 parts of dispersant and 0.2-0.5 parts of coupling agent.

[0050] Especially, the PBT particle realizes high glowing wire and low smoke toxicity through the halogen-based flame retardant, the nitrogen-based MCA flame retardant and the auxiliary additive, and ensures that the formula design takes into account the performance of high glowing wire and high impact and the environmental protection characteristics.

[0051] Compared with the prior art, the application has the beneficial effects that through the multi-model collaborative architecture of CNN, attention mechanism and LSTM, the application realizes the short-term key features of the actual feeding speed of the composite flame retardant and glass fiber in the PBT particle preparation process, identifies the characteristic signal of bridging and bridging, and then maps and generates the feeding speed that can actively predict the feeding fluctuation and inhibit the bridging and bridging, thereby ensuring the uniformity of the component dispersion and realizing the preparation performance stability of the environmentally friendly PBT particle with high glowing wire and high impact.

[0052] Especially, the application is based on intelligent feeding control of multi-scale CNN, attention mechanism, LSTM and full connection network, solves the problems of proportion imbalance caused by powder caking and fiber winding in traditional feeding, and realizes the following effects: the double-branch multi-scale CNN combines the attention mechanism, accurately captures the powder caking trend, flow stability and fiber winding risk, and fracture characteristics, and through attention weighting, improves the feature recognition accuracy; the fast and slow double-branch LSTM processes the powder instantaneous mutation and fiber long-term trend, the full connection network outputs the prediction speed and target value deviation control, realizes the advance active correction, ensures the component dispersion uniformity of the feeding, and greatly improves the batch consistency of product quality.

[0053] Especially, the PBT particles of the application realize high glow wire and low smoke toxicity through halogen flame retardant, nitrogen MCA flame retardant and auxiliary additive, and ensure that the formula design considers the performance of high glow wire and high impact and environmental protection characteristics. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 It is a flowchart of the environmentally friendly high-glow-wire high-impact PBT particles and the preparation method thereof of the embodiment of the application;

[0055] Figure 2 It is a powder feature extraction branch flowchart of the environmentally friendly high-glow-wire high-impact PBT particles and the preparation method thereof of the embodiment of the application;

[0056] Figure 3 It is an LSTM model flowchart of the environmentally friendly high-glow-wire high-impact PBT particles and the preparation method thereof of the embodiment of the application;

[0057] Figure 4 It is a preparation overall flowchart of the environmentally friendly high-glow-wire high-impact PBT particles of the embodiment of the application. DETAILED DESCRIPTION

[0058] In order to make the purpose and advantages of the application more clear and explicit, the application will be further described below in combination with embodiments; it should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

[0059] The preferred embodiments of the application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the application, and are not used to limit the protection scope of the application.

[0060] It should be noted that in the description of the present application, the terms "upper", "lower", "left", "right", "inner", "outer" and the like indicate the direction or positional relationship of the terms based on the direction or positional relationship shown in the drawings, which is only for the convenience of description, and does not indicate or imply that the device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0061] In addition, it should be noted that in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0062] As shown in Figures 1 to 4 The present application provides an environmentally friendly high-burning wire high-impact PBT particle and a preparation method thereof. Through the multi-model collaborative architecture of CNN, attention mechanism and LSTM, the short-term key features of the actual feeding speed of the composite flame retardant and glass fiber in the preparation process of the PBT particle are realized, the characteristic signals of bridging and caking are identified, and then the feeding speed capable of actively predicting feeding fluctuation and inhibiting bridging and caking is generated, thereby ensuring the uniformity of component dispersion and realizing the stable preparation performance of the environmentally friendly high-burning wire high-impact PBT particle.

[0063] As shown in Figure 1 The present embodiment provides a preparation method of an environmentally friendly high-burning wire high-impact PBT particle. The component composition of the PBT particle includes PBT resin, glass fiber and composite flame retardant. The preparation method of the PBT particle includes:

[0064] The composite flame retardant powder and the glass fiber are added to the forced feeding equipment, and the PBT resin is forcedly fed, and then the PBT particle premix is obtained. The PBT particle is obtained by extrusion injection molding of the PBT particle premix;

[0065] The forced feeding equipment obtains the actual feeding speed of the flame retardant powder, the actual feeding speed of the glass fiber and the equipment discharge speed through a plurality of loss-in-weight scales respectively;

[0066] The actual feeding speed of the flame retardant powder, the actual feeding speed of the glass fiber and the equipment discharge speed are subjected to short-term feature extraction by a feature extraction model based on a multi-scale CNN and an attention mechanism architecture to generate a feeding feature vector;

[0067] The feeding characteristics vector is captured by an LSTM model to generate a feeding bridge block characteristic;

[0068] The feeding bridge block characteristic is passed through a full connection network to generate a predicted powder feeding speed and a predicted fiber feeding speed, and the expected powder feeding speed and the expected fiber feeding speed of the forced feeding device are adjusted based on the predicted powder feeding speed and the predicted fiber feeding speed.

[0069] In particular, the feature extraction model using a multi-scale CNN (Convolutional Neural Network) and an attention mechanism architecture can respectively extract short-term feeding characteristics in line with the characteristics of the composite flame retardant powder, which is prone to agglomeration and has large flow variability, and the glass fiber, which is prone to winding and has uniform density. LSTM can respectively extract long-term feeding bridge block characteristics in line with the characteristics of the feeding process of the composite flame retardant powder, which has unstable flow, and the glass fiber, which needs continuous and smooth feeding. Then, the feeding bridge block characteristics extracted by the output mapping model of the MLP architecture can map to generate a predicted powder feeding speed that is more in line with the composite flame retardant powder and pays more attention to fluctuations in a short period of time, and a predicted fiber feeding speed that is more in line with the glass fiber and pays more attention to stability in a long period of time. Further, the feeding speed adjustment of the forced feeding device realizes more accurate prediction control, reduces the probability of bridge formation and agglomeration, ensures the uniform dispersion of the glass fiber and the composite flame retardant and the main material PBT tree mixed in the forced feeding device, and realizes stable performance of the environmentally friendly high-glowing and high-impact PBT particle product.

[0070] Further, the feature extraction model includes a powder feature extraction branch and a fiber feature extraction branch, the feeding characteristic vector includes a powder feeding characteristic vector and a fiber feeding characteristic vector, and the process of generating the feeding characteristic vector by short-term feature extraction by the feature extraction model includes:

[0071] The actual powder feeding speed, the powder target feeding speed, and the feeding machine actual running data are taken as powder branch input data, and the powder branch input data is passed through the powder multi-scale convolution layer of the powder feature extraction branch to generate a multi-scale powder feeding characteristic;

[0072] The multi-scale powder feeding characteristic and the powder branch input data are guided and weighted by the powder attention mechanism of the powder feature extraction branch to generate the powder feeding characteristic vector;

[0073] The actual fiber feeding speed, the fiber target feeding speed, and the feeding machine actual running data are taken as fiber branch input data, and the fiber branch input data is passed through the fiber multi-scale convolution layer of the fiber feature extraction branch to generate a multi-scale fiber feeding characteristic;

[0074] The multi-scale fiber feeding feature and fiber branch input data are multi-scale feature guided weighted through the fiber feature extraction branch fiber attention mechanism to generate the fiber feeding feature vector.

[0075] Preferably, the feeder actual operation data includes motor speed of the feeding mixture and device discharge speed.

[0076] In particular, the powder branch focuses on the caking and bridging characteristics of the powder, and the fiber branch adapts to the winding and breaking characteristics of the fiber, avoiding the confusion of powder and fiber features.

[0077] Further, the powder multi-scale convolution layer includes a first powder convolution layer, a second powder convolution layer, and a third powder convolution layer, the multi-scale powder feeding feature includes powder fluctuation characteristics, powder bridging trend characteristics, and powder flow characteristics, and the process of generating a powder feeding feature vector through a powder multi-scale convolution layer and a powder attention mechanism includes:

[0078] The powder branch input data is respectively passed through the first powder convolution layer, the second powder convolution layer, and the third powder convolution layer to generate the powder fluctuation characteristics, the powder bridging trend characteristics, and the powder flow characteristics;

[0079] The standard deviation, skewness, and kurtosis of the powder branch input data are respectively passed through the powder attention mechanism to generate fluctuation attention weights, caking attention weights, and bridging attention weights, and the fluctuation attention weights, caking attention weights, and bridging attention weights are weighted and summed to the powder fluctuation characteristics, powder bridging trend characteristics, and powder flow characteristics to generate the powder feeding feature vector in turn;

[0080] The fiber multi-scale convolution layer includes a first fiber convolution layer, a second fiber convolution layer, and a third fiber convolution layer, the multi-scale fiber feeding feature includes fiber breaking characteristics, fiber winding trend characteristics, and fiber conveying characteristics, and the process of generating a fiber feeding feature vector through a fiber multi-scale convolution layer and a fiber attention mechanism includes:

[0081] The fiber branch input data is respectively passed through the first fiber convolution layer, the second fiber convolution layer, and the third fiber convolution layer to generate the fiber breaking characteristics, the fiber winding trend characteristics, and the fiber conveying characteristics in turn;

[0082] The mean, variance and information entropy of the fiber branch input data are input into the fiber attention mechanism respectively to generate fracture attention weights, winding attention weights and feeding attention weights, and the fracture attention weights, winding attention weights and feeding attention weights are weighted and summed on the fiber fracture feature, fiber winding trend feature and fiber feeding feature to generate the fiber feeding feature vector;

[0083] Wherein, the convolution kernel size of the convolution layer is in order from small to large: the first powder convolution layer, the first fiber convolution layer, the second powder convolution layer, the second fiber convolution layer, the third powder convolution layer and the third fiber convolution layer.

[0084] Specifically, the powder multi-scale convolution layer can be represented as:

[0085]

[0086] In the formula, representing the powder fluctuation feature, the powder bridging trend feature and the powder flow feature respectively, representing activation function, representing the first powder convolution layer 3x1 size convolution kernel, the second powder convolution layer 7x1 size convolution kernel and the third powder convolution layer 15x1 size convolution kernel of the powder feature extraction branch p respectively, which cover the feeding data within 0.6 seconds, 1.4 seconds and 3 seconds respectively, representing the powder branch input data of the powder feature extraction branch p, the powder branch input data including the actual feeding speed of the flame-retardant powder, the target feeding speed of the powder and the actual running data of the feeder, representing the 3x1 convolution kernel bias term of the first powder convolution layer, the 7x1 convolution kernel bias term of the second powder convolution layer and the 15x1 convolution kernel bias term of the third powder convolution layer of the powder feature extraction branch p respectively.

[0087] Especially, the powder multi-scale convolution layer realizes multi-scale feature extraction for the characteristics of the composite flame-retardant powder prone to agglomeration.

[0088] Specifically, the powder attention mechanism can be represented as:

[0089]

[0090] In the formula, representing the attention weight vector of the powder attention mechanism p, containing 3 elements, representing normalization function, representing the learnable weight matrix and the learnable bias term of the powder attention mechanism p respectively, respectively represent the standard deviation, skewness and kurtosis of the powder branch input data, the powder branch input data including the actual feeding speed of the fire-retardant powder, the target feeding speed of the powder and the actual running data of the feeder, wherein the standard deviation reflects the fluctuation degree of the powder feeding data by statistically analyzing the dispersion of the actual feeding speed of the fire-retardant powder, the skewness reflects the typical precursor degree of caking by statistically analyzing the slow recovery after the short-term sudden drop of the powder feeding speed, and the kurtosis can reflect the instantaneous impact of the caking through the feeder, represents the powder feeding characteristic vector, respectively represent the three elements of the attention weight vector, namely the fluctuation attention weight, the caking attention weight and the bridging attention weight, respectively represent the powder fluctuation feature, the powder bridging trend feature and the powder flow feature.

[0091] In particular, the powder attention mechanism realizes that different importance weights are assigned to different features reflected in the powder input data, so that the feature extraction model learns to pay attention to more important information.

[0092] Specifically, the powder multi-scale convolution layer can be represented as:

[0093]

[0094] In the formula, respectively represent the fiber breakage feature, the fiber winding trend feature and the fiber conveying feature of the fiber feature extraction branch f, represents an activation function, respectively represent the first fiber convolution layer 5x1 size convolution kernel, the second fiber convolution layer 11x1 size convolution kernel and the third fiber convolution layer 21x1 size convolution kernel of the fiber feature extraction branch f, which cover the feeding data within 0.6 seconds, 1.4 seconds and 3 seconds respectively, represents the fiber branch input data of the fiber feature extraction branch f, the fiber branch input data including the actual feeding speed of the glass fiber, the target feeding speed of the fiber and the actual running data of the feeder, respectively represent the 5x1 convolution kernel bias term of the first fiber convolution layer, the 11x1 convolution kernel bias term of the second powder fiber layer and the 21x1 convolution kernel bias term of the third fiber convolution layer of the fiber feature extraction branch f.

[0095] Specifically, the fiber attention mechanism can be represented as:

[0096]

[0097] In the formula, represents the attention weight vector of the fiber attention mechanism f, containing 3 elements, represents a normalization function, respectively represent the learnable weight matrix and the learnable bias term of the fiber attention mechanism f, respectively represent the mean, variance and information entropy of the fiber branch input data, the fiber branch input data including the actual fiber feeding speed, the target fiber feeding speed and the actual feeding machine running data, wherein the mean measures the average rate stability of the fiber feeding to reflect whether the fiber feeding is not smooth due to winding, the variance measures the fluctuation degree of the fiber feeding rate to reflect whether the fiber has local winding or breakage, and the information entropy measures the disorder of the fiber feeding data, which is calculated by the information entropy classical formula of discrete probability as wherein represents the probability of data falling into the interval k, represents the fiber feeding feature vector, respectively represent the three elements of the attention weight vector, i.e. the breakage attention weight, the winding attention weight and the feeding attention weight, respectively represent the fiber breakage feature, the fiber winding trend feature and the fiber conveying feature of the fiber feature extraction branch f.

[0098] In particular, by the powder attention mechanism, different importance weights are assigned to different features reflected in the fiber input data, so that the feature extraction model learns to pay attention to more important information.

[0099] Further, the LSTM model includes a powder fast LSTM branch, a fiber slow LSTM branch and an interaction layer, and the process of generating the feeding bridge caking feature by the LSTM model includes:

[0100] The powder feeding feature vector and its change amount are respectively passed through the powder fast LSTM branch with a fiber interaction input gate and a change rate enhanced forgetting gate to generate a powder hidden state;

[0101] The fiber feeding feature vector and its change amount are respectively passed through the fiber slow LSTM branch with a trend enhanced input gate and a smoothing forgetting gate to generate a fiber hidden state;

[0102] The powder hidden state and the fiber hidden state are passed through the interaction layer to generate the feeding bridge caking feature.

[0103] In particular, in view of the time dynamic difference and material coupling characteristics of the powder and glass fiber feeding abnormalities, by branch function differentiation and interaction fusion, comprehensive capture of the time sequence characteristics of the complex feeding system is realized, accurate modeling of the instantaneous evolution of powder caking, the development of fiber winding and the linkage of abnormal materials is realized, and high-accuracy feature support is provided for feeding bridge caking early warning and dynamic regulation.

[0104] Further, the process of generating the powder hidden state comprises:

[0105] linearly mapping the concatenation vector of the powder hidden state of the last time step, the powder feeding feature vector and the fiber hidden state of the last time step respectively to generate the fiber interaction input gate and the initial forgetting gate, performing weighted summation on the difference of the powder feeding feature vectors of multiple time steps on the initial forgetting gate to generate the change rate enhanced forgetting gate, and updating the powder hidden state of the current time step based on the fiber interaction input gate and the change rate enhanced forgetting gate;

[0106] The process of generating the fiber hidden state comprises:

[0107] calculating a glass fiber feeding trend value based on the difference between the actual glass fiber feeding speed and the fiber target feeding speed within a set time window, performing weighted summation on the initial input gate based on the fiber feeding feature vector and the glass fiber feeding trend value to generate the trend enhanced input gate, performing weighted addition on the fiber initial forgetting gate based on the fiber feeding feature vector and the change rate of the fiber feeding feature vector to generate the smooth forgetting gate, and updating the fiber hidden state of the current time step based on the trend enhanced input gate and the smooth forgetting gate.

[0108] In particular, the fine-grained LSTM architecture dynamically adapts through the gate mechanism, fuses cross-material information, and enhances time sequence features, so that the double branches can more accurately capture the time sequence feature rules of the respective materials, while strengthening cross-material correlation perception.

[0109] Specifically, the powder fast LSTM branch can be represented as:

[0110]

[0111] In the formula, respectively represent the fiber interaction input gate, the change rate enhanced forgetting gate, the output gate, the candidate cell state, the current cell state update, the last time step cell state update and the powder hidden state of the current time step of the powder fast LSTM branch, represents the Sigmoid function, represents function, represents element-wise multiplication, respectively represent the weight matrix of the input gate, the forgetting gate, the output gate and the candidate cell state of the powder fast LSTM branch, respectively represent the bias term of the input gate, the forgetting gate, the output gate and the candidate cell state of the powder fast LSTM branch, represents the last time step powder hidden state the powder feeding feature vector of the t-th time step the fiber hidden state of the previous time step the concatenation vector, denotes a change rate weighting coefficient, wherein denotes an initial input gate, denotes a change rate of the powder feeding feature vector of the plurality of time steps, preferably, the pairwise difference of the closest three time steps is averaged.

[0112] Specifically, the fiber fast LSTM branch can be represented as:

[0113]

[0114] In the formula, respectively denote a trend enhancement input gate, a smooth forgetting gate, an output gate, a candidate cell state, a current cell state update, a previous time step cell state update, a previous previous time step cell state update, a fiber hidden state of the current time step, and a fiber hidden state of the previous time step of the fiber fast LSTM branch, denotes a Sigmoid function, denotes a function, denotes element-wise multiplication, respectively denote a weight matrix of an input gate, a forgetting gate, an output gate, and a candidate cell state of the fiber fast LSTM branch, respectively denote a bias term of an input gate, a forgetting gate, an output gate, and a candidate cell state of the fiber fast LSTM branch, and [ ] denotes a feature vector concatenation operation, respectively denote a trend enhancement weighting coefficient, a change weakening weighting coefficient, and a historical cell state retention coefficient, denotes a glass fiber feeding trend value, denotes a change rate of a fiber feeding feature vector , wherein denotes an initial input gate, denotes a fiber initial forgetting gate.

[0115] The calculation process of the glass fiber feeding trend value can be represented as:

[0116]

[0117] In the formula, denotes a glass fiber feeding trend value, denotes a set time window, preferably 5, denotes an actual glass fiber feeding speed, denotes a target fiber feeding speed.

[0118] Further, the process of generating the feeding bridge caking feature through the interaction layer includes:

[0119] weighting sum of the powder hidden state and the fiber hidden state to generate a linear term;

[0120] element-wise multiplication of the powder hidden state and the fiber hidden state to generate an interaction term;

[0121] weighting sum of the linear term and the interaction term to generate a comprehensive hidden state, and taking the comprehensive hidden state, the powder hidden state and the fiber hidden state of the final time step as the feeding bridge caking feature.

[0122] In particular, the hidden states output by the powder fast LSTM branch and the fiber slow LSTM branch are deeply fused, and the finally generated feeding bridge caking feature has information of powder instantaneous abnormal details, fiber long-term trend rules and cross-material linkage characteristics, thereby providing high-recognizability feature support for feeding dynamic regulation.

[0123] Specifically, the interaction layer can be represented as:

[0124]

[0125] In the formula, comprehensive hidden state of the current time step, indicates a set time window, preferably 5, actual glass fiber feeding speed, target fiber feeding speed.

[0126] Specifically, the feeding bridge caking feature is passed through a fully connected network to generate a predicted powder feeding speed and a predicted fiber feeding speed, and the fully connected network includes a ReLU activation function convolution layer, a Dropout layer, a residual connection layer and a LayerNorm layer.

[0127] Further, the process of adjusting the expected powder feeding speed and the expected fiber feeding speed includes:

[0128] based on the difference between the predicted powder feeding speed and the powder target feeding speed to generate a powder adjustment amount, and based on the calculation of the powder adjustment amount and the current powder feeding speed to generate an expected powder feeding speed;

[0129] based on the difference between the predicted fiber feeding speed and the fiber target feeding speed to generate a fiber adjustment amount, and based on the calculation of the fiber adjustment amount and the current fiber feeding speed to generate an expected fiber feeding speed.

[0130] In particular, the closed-loop control mechanism of the feeding speed avoids the limitations of traditional fixed speed or empirical adjustment, and realizes active dynamic and accurate regulation of the feeding speed of the feeding machine.

[0131] Specifically, the process of generating the expected powder feeding speed and the expected fiber feeding speed can be represented as:

[0132]

[0133] wherein, respectively represent the expected powder feeding speed and the expected fiber feeding speed at the next moment, respectively represent the conversion coefficients of the two feeding speeds and the motor speed, respectively represent the conversion coefficients of the two feeding speeds and the motor speed, respectively represent the predicted powder feeding speed and the predicted fiber feeding speed, respectively represent the powder target feeding speed and the fiber target feeding speed, respectively represent the current powder feeding speed and the current fiber feeding speed.

[0134] Further, the preparation method further comprises:

[0135] constructing a main loss term based on the mean square error of the predicted powder feeding speed and the sample true powder optimal feeding speed, and the mean square error of the predicted fiber feeding speed and the sample true fiber optimal feeding speed;

[0136] constructing a flow stability loss term based on the ratio of the predicted powder feeding speed and the predicted fiber feeding speed, and the ratio of the powder theoretical flow and the fiber theoretical flow;

[0137] constructing a comprehensive loss function based on the weighted sum of the main loss term and the flow stability loss term, and performing collaborative optimization training on the feature extraction model, the LSTM model and the full connection network through the comprehensive loss function.

[0138] In particular, the flow stability loss term is constructed by the deviation of the ratio of the predicted powder and fiber speeds and the ratio of the powder and fiber theoretical flow (the formula requirement), which forces the model to maintain the material ratio within the formula requirement range and keep stable (without caking and bridging, etc.) during prediction, avoids product quality problems caused by unbalanced ratio, and ensures the product quality of forced feeding.

[0139] Specifically, the process of generating the expected powder feeding speed and the expected fiber feeding speed can be represented as:

[0140]

[0141] wherein, represents the comprehensive loss function, and N represents the total number of samples, respectively represent the weighted coefficients of the main loss term and the flow stability loss term, respectively represent the predicted powder feeding speed and the predicted fiber feeding speed of the i-th input sample, respectively represent the sample real powder optimal feeding speed and the sample real fiber optimal feeding speed of the i-th input sample, respectively represent the bulk densities of the composite flame retardant and the glass fiber, which are determined according to the formula ratio of the PTB particles, respectively represent the cross-sectional areas of the composite flame retardant and the glass fiber feeding ports of the forced feeding device.

[0142] The embodiment also provides an environmentally-friendly high-glow-wire high-impact PBT particle prepared by the preparation method, and the PBT particle comprises, in parts by weight, 30-50 parts of PBT resin, 5-40 parts of glass fiber and 23.9-60.4 parts of composite flame retardant, wherein the composite flame retardant comprises a halogen-based flame retardant, a nitrogen-based MCA flame retardant and an auxiliary additive.

[0143] The halogen-based flame retardant comprises, in parts by weight, 8-15 parts of brominated epoxy, 5-15 parts of brominated polystyrene and 1-5 parts of antimony trioxide.

[0144] The nitrogen-based MCA flame retardant comprises, in parts by weight, 3-10 parts of nitrogen-based flame retardant MCA, 2-6 parts of stannate and 1-3 parts of silicate.

[0145] In particular, in the original high-glow-wire PBT formula, the triphenyl phosphate TPP and the nitrogen-based flame retardant need to be compounded in the bromine-antimony flame retardant system to form carbon and gas-phase oxygen insulation, so that the PBT does not produce open fire under the condition of glow-wire 750 degrees. However, it is currently considered that the triphenyl phosphate TPP is not environmentally friendly and is harmful to the human body, and therefore the PBT particle of the embodiment realizes the environmentally-friendly high-glow-wire high-impact performance without triphenyl phosphate TPP.

[0146] Further, the auxiliary additive comprises, in parts by weight, 3-5 parts of a toughening agent, 0.2-0.4 parts of an antioxidant, 0.5 parts of a dispersant and 0.2-0.5 parts of a coupling agent.

[0147] Specifically, the process of preparing the composite flame retardant powder from the halogen-based flame retardant, the nitrogen-based MCA flame retardant and the auxiliary additive comprises: drying the PBT resin, the halogen-based flame retardant, the nitrogen-based MCA flame retardant and the auxiliary additive in an oven at 110 degrees for 3 hours, and testing the moisture content with a moisture meter to keep the moisture content below 0.1%; stirring the coupling agent and the flame retardant powder in a powder mill to achieve pre-dispersion; and fully mixing and stirring all the materials of the halogen-based flame retardant, the nitrogen-based MCA flame retardant and the auxiliary additive to form the composite flame retardant powder.

[0148] The composite flame retardant powder, PBT resin and glass fiber are made into PBT particle premix through forced feeding equipment. The process of preparing PBT particles from PBT particle premix includes: PBT particle premix is sequentially extruded through a double screw extruder with a temperature of 225-255 degrees, stretched through water (cold water tank), dried (water blowing and water suction machine), granulated (granulator), screened (vibrating screen), dried (oven), injection molded (injection molding machine and mold), and tested (hot wire tester and impact tester) to make PBT particles.

[0149] Specifically, the test effect of the components of the PBT particles is shown in the following table "PBT flame-retardant reinforced finished product test specification":

[0150] In particular, the PBT particles achieve high glow wire and low smoke toxicity through halogen-based flame retardants, nitrogen-based MCA flame retardants and auxiliary additives, ensuring that the formula design takes into account both high glow wire and high impact performance and environmental characteristics.

[0151] In this embodiment, through the multi-model collaborative architecture of CNN, attention mechanism and LSTM, the short-term key features based on the actual feeding speed of the composite flame retardant and glass fiber in the preparation process of the PBT particles are realized, the characteristic signals of bridging and caking are identified, and then the feeding speed capable of actively predicting feeding fluctuations and inhibiting bridging and caking is generated, thereby ensuring the uniformity of component dispersion and realizing the stable preparation performance of the environmentally friendly high-glow-wire high-impact PBT particles. The intelligent feeding control based on multi-scale CNN, attention mechanism, LSTM and fully connected network solves the problem of proportion imbalance caused by powder caking and fiber winding in traditional feeding. The fast and slow double-branch LSTM processes the instantaneous mutation of the powder and the long-term trend of the fiber, and the deviation control of the predicted speed output by the fully connected network and the target value, realizes the advance active correction, ensures the uniformity of component dispersion of the feeding, and greatly improves the batch consistency of product quality. The PBT particles achieve high glow wire and low smoke toxicity through halogen-based flame retardants, nitrogen-based MCA flame retardants and auxiliary additives, ensuring that the formula design takes into account both high glow wire and high impact performance and environmental characteristics.

[0152] Those skilled in the art can appreciate that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0153] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will all fall within the protection scope of the present application.

[0154] The above description is only the preferred embodiments of the present application and is not intended to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for preparing environmentally friendly high-heat-wire, high-impact PBT particles, characterized in that, include: Composite flame retardant powder and glass fiber are added to a forced feeding device to force-feed PBT resin, and then PBT granule premix is ​​discharged. The PBT granule premix is ​​extruded and injection molded to obtain PBT granules. The forced feeding device obtains the actual feeding speed of flame retardant powder, the actual feeding speed of glass fiber, and the equipment discharge speed through multiple loss-in-weight scales. The actual feeding speed of the flame retardant powder, the actual feeding speed of the glass fiber, and the discharge speed of the equipment are used to perform short-term feature extraction using a feature extraction model based on a multi-scale CNN and attention mechanism architecture to generate a feeding feature vector. The feeding feature vector is used to capture the feeding pattern through an LSTM model to generate feeding bridging and agglomeration features; The feed bridging and agglomeration characteristics are passed through a fully connected network to generate predicted powder feed rate and predicted fiber feed rate, and the desired powder feed speed and desired fiber feed speed of the forced feeder are adjusted based on the predicted powder feed rate and predicted fiber feed rate.

2. The method for preparing environmentally friendly high-heat-wire, high-impact PBT particles according to claim 1, characterized in that, The feature extraction model includes a powder feature extraction branch and a fiber feature extraction branch. The feeding feature vector includes a powder feeding feature vector and a fiber feeding feature vector. The process of generating the feeding feature vector through short-term feature extraction using the feature extraction model includes: The actual feeding speed of the flame retardant powder, the target feeding speed of the powder, and the actual operating data of the feeder are used as the powder branch input data. The powder branch input data is then passed through the powder multi-scale convolutional layer of the powder feature extraction branch to generate multi-scale powder feeding features. The multi-scale powder feeding features and the powder branch input data are weighted by multi-scale feature guidance through the powder attention mechanism of the powder feature extraction branch to generate the powder feeding feature vector. The actual feeding speed of the glass fiber, the target feeding speed of the fiber, and the actual operating data of the feeder are used as fiber branch input data. The fiber branch input data are then processed through the fiber feature extraction branch multi-scale convolutional layer to generate multi-scale fiber feeding features. The multi-scale fiber feeding features and fiber branch input data are weighted by multi-scale feature guidance through the fiber attention mechanism of the fiber feature extraction branch to generate the fiber feeding feature vector.

3. The method for preparing environmentally friendly high-heat-wire, high-impact PBT particles according to claim 2, characterized in that, The multi-scale powder feeding characteristics include powder fluctuation characteristics, powder bridging trend characteristics, and powder flow characteristics. The process of generating the powder feeding feature vector through multi-scale convolutional layers and powder attention mechanisms includes: The powder branch input data is passed through the first powder convolutional layer, the second powder convolutional layer and the third powder convolutional layer of the powder multi-scale convolutional layer respectively to generate powder fluctuation characteristics, powder bridging trend characteristics and powder flow characteristics in sequence; The standard deviation, skewness, and kurtosis of the powder branch input data are respectively processed through the powder attention mechanism to generate fluctuation attention weight, agglomeration attention weight, and bridging attention weight. The fluctuation attention weight, agglomeration attention weight, and bridging attention weight are then weighted and summed with the powder fluctuation characteristics, powder bridging trend characteristics, and powder flow characteristics to generate the powder feeding feature vector. The multi-scale fiber feeding features include fiber breakage features, fiber entanglement trend features, and fiber transport features. The process of generating fiber feeding feature vectors through fiber multi-scale convolutional layers and fiber attention mechanisms includes: The fiber branch input data is passed through the first fiber convolutional layer, the second fiber convolutional layer, and the third fiber convolutional layer of the fiber multi-scale convolutional layer respectively to generate fiber breakage characteristics, fiber entanglement trend characteristics, and fiber transport characteristics in sequence. The mean, variance, and information entropy of the fiber branch input data are respectively processed through the fiber attention mechanism to generate breakage attention weight, entanglement attention weight, and feeding attention weight. The breakage attention weight, entanglement attention weight, and feeding attention weight are then weighted and summed with the fiber breakage feature, fiber entanglement trend feature, and fiber transport feature to generate the fiber feeding feature vector. The convolution kernel sizes of the convolutional layers, from smallest to largest, are as follows: the first powder convolutional layer, the first fiber convolutional layer, the second powder convolutional layer, the second fiber convolutional layer, the third powder convolutional layer, and the third fiber convolutional layer.

4. The method for preparing environmentally friendly high-heat-wire, high-impact PBT particles according to claim 2, characterized in that, The LSTM model includes a fast LSTM branch for powders, a slow LSTM branch for fibers, and an interaction layer. The process of generating feed bridging and agglomeration features using the LSTM model includes: The powder feeding feature vector and its change are respectively passed through a fast LSTM branch for powder with fiber interaction input gate and rate of change enhancement forget gate to generate the powder hidden state; The fiber feeding feature vector and its changes are passed through a slow LSTM branch with a trend-enhancing input gate and a smooth forgetting gate to generate the fiber hidden state. The powder hidden state and fiber hidden state are passed through an interaction layer to generate the feed bridging agglomeration feature.

5. The method for preparing environmentally friendly high-heat-wire, high-impact PBT particles according to claim 4, characterized in that, The process of generating a hidden state for powder includes: The powder hidden state, the powder feeding feature vector, and the spliced ​​vector of the fiber hidden state from the previous time step are linearly mapped to generate the fiber interaction input gate and the initial forget gate. The differences of the powder feeding feature vectors from multiple time steps are weighted and summed on the initial forget gate to generate the rate of change enhanced forget gate. The powder hidden state at the current time step is updated based on the fiber interaction input gate and the rate of change enhanced forget gate. The process of generating the hidden state of fibers includes: The glass fiber feeding trend value is calculated based on the difference between the actual feeding speed and the target feeding speed of the glass fiber within a set time window. The initial input gate based on the fiber feeding feature vector and the glass fiber feeding trend value are weighted and summed to generate the trend enhancement input gate. The initial forgetting gate based on the fiber feeding feature vector is weighted and summed with the rate of change of the fiber feeding feature vector to generate the smooth forgetting gate. The hidden state of the fiber at the current time step is updated based on the trend enhancement input gate and the smooth forgetting gate.

6. The method for preparing environmentally friendly high-heat-wire, high-impact PBT particles according to claim 4, characterized in that, The process of generating feed bridging cluster features through the interaction layer includes: The powder hidden state and the fiber hidden state are weighted and summed to generate a linear term; The interaction term is obtained by multiplying the powder hidden state and the fiber hidden state element by element. The linear term and the interaction term are weighted and summed to generate a comprehensive hidden state, and the comprehensive hidden state, the powder hidden state, and the fiber hidden state at the final time step are used as the feed bridging and agglomeration features.

7. The method for preparing environmentally friendly high-heat-wire, high-impact PBT particles according to claim 2, characterized in that, The process of adjusting the desired powder feed speed and the desired fiber feed speed includes: Based on the difference between the predicted powder feeding speed and the target powder feeding speed, a powder adjustment amount is generated. Based on the powder adjustment amount and the calculation of the current powder feeding speed, a desired powder feeding speed is generated. The fiber adjustment amount is generated based on the difference between the predicted fiber feeding speed and the target fiber feeding speed, and the desired fiber feeding speed is generated based on the calculation of the fiber adjustment amount and the current fiber feeding speed.

8. The method for preparing environmentally friendly high-glow wire high-impact PBT particles according to any one of claims 1 to 7, characterized in that, Also includes: Based on the mean square error between the predicted powder feeding rate and the optimal powder feeding rate of the sample, and the mean square error between the predicted fiber feeding rate and the optimal fiber feeding rate of the sample, a main loss term is constructed. A flow stability loss term is constructed based on the ratio of the predicted powder feeding rate to the predicted fiber feeding rate and the ratio of the theoretical powder flow rate to the theoretical fiber flow rate. A comprehensive loss function is constructed based on the weighted sum of the main loss term and the traffic stability loss term. The feature extraction model, the LSTM model, and the fully connected network are then collaboratively optimized and trained using the comprehensive loss function.

9. An environmentally friendly high-glow wire, high-impact PBT particle prepared by the method described in any one of claims 1 to 8, characterized in that, By weight, it comprises 30-50 parts PBT resin, 5-40 parts glass fiber and 23.9-60.4 parts composite flame retardant, wherein the composite flame retardant includes halogenated flame retardant, nitrogen-based MCA flame retardant and auxiliary additives. The halogenated flame retardant, by weight, comprises 8-15 parts brominated epoxy, 5-15 parts brominated polystyrene and 1-5 parts antimony trioxide; The nitrogen-based MCA flame retardant comprises, by weight, 3-10 parts nitrogen-based flame retardant MCA, 2-6 parts stannate and 1-3 parts silicate.

10. The environmentally friendly high-heat-wire, high-impact PBT granules according to claim 9, characterized in that, The auxiliary additives, by weight, include 3-5 parts toughening agent, 0.2-0.4 parts antioxidant, 0.5 parts dispersant and 0.2-0.5 parts coupling agent.

Images