Ignition charge-free electronic detonator and quality prediction and control method thereof

Abstract

The invention discloses an ignition charge-free electronic detonator and a quality prediction and control method thereof. The detonator structurally comprises a wire clamp, an injection molding plug, an electronic control module and a basic detonator, a leg wire for connecting the electric ignition head and the exploder is arranged in the injection molding plug; an elastic piece is arranged on the surface of the electronic control module. The method comprises the following steps: S1, collecting production raw material data of the non-ignition charge electronic detonator; s2, cleaning and standardizing the collected data; s3, constructing a feature extraction model and a special prediction model, and outputting a sensitivity index prediction result and quality risk early warning; and S4, according to a large model output result, executing a subsequent production process on a qualified batch, and carrying out secondary analysis on a risk batch. Complete support is provided by collecting production whole process quality related data, the abnormal value risk weight is marked in combination with historical fault reworking data to improve the effectiveness of the model, risks can be solved in advance, waste is reduced, and the quality stability and production efficiency of the ignition charge-free electronic detonator are improved.

Description

Technical Field

[0001] This invention relates to the field of civil explosives technology, specifically to an electronic detonator without ignition charge and its quality prediction and control method. Background Technology

[0002] Existing industrial electric detonators and industrial electronic detonators all use a sensitive igniter mixed with adhesive to form a spherical igniter head, which is then coated and dried using a heating resistor. The igniter head is then inserted into a base detonator and connected via a bayonet device to create an industrial detonator.

[0003] To convert to a non-ignition electronic detonator, it is necessary to address issues in the quality control of non-ignition detonators, such as incomplete data coverage, lack of risk correlation in outlier handling, insufficient prediction accuracy, and non-closed-loop control processes. Current processes often fail to fully collect key quality-related data from the entire production process, including raw materials, manufacturing processes, and testing, resulting in a lack of comprehensive data support for quality analysis. Outlier handling in the collected data often involves simple removal or a uniform approach, without considering historical failure and rework data to prioritize different anomalies and differentiate the severity of their impact on quality.

[0004] Meanwhile, the lack of relevant means to process the accuracy of sensitivity index prediction results leads to delayed quality risk warnings or a high rate of misjudgment. Moreover, after risks are discovered, they are mostly dealt with after the fact, making it impossible to adjust process parameters in a timely manner and verify the adjustment effect through secondary data collection. This results in low rework efficiency for non-conforming batches, serious waste of resources, and difficulty in consistently ensuring product quality. Summary of the Invention

[0005] This invention provides a ignition-free electronic detonator and a method for predicting and controlling its quality, in order to solve the above-mentioned problems.

[0006] To solve the above technical problems, the present invention provides a ignition-free electronic detonator, comprising: Line clamps, injection plugs, electronic control modules, and basic detonators; The injection plug is installed at the opening of the base detonator and contains lead wires for connecting the electric ignition head and the detonator. The electronic control module is located inside the base detonator, and its surface is equipped with spring contacts for communication with external circuits.

[0007] As an optional method, the basic detonator also contains an initiating explosive and a main explosive charge, with the injection-molded plug welded to the electronic control module.

[0008] On the other hand, the present invention also provides a method for predicting and controlling the quality of an ignition-free electronic detonator, comprising: S1. Collect raw material data, production process data and test data related to quality throughout the entire production process of ignition-free electronic detonators; S2. The collected data is cleaned and standardized. At the same time, outlier weights are labeled based on historical fault rework data to generate a training sample set with risk priority. S3. Construct a feature extraction model and a special prediction model, input the data preprocessed in S2 above, and output the sensitivity index prediction results and quality risk warning of the electronic detonator without ignition powder. S4. Based on the output of the large model, execute subsequent production processes for qualified batches, and adjust the process parameters for risky batches before re-collecting data and performing secondary analysis until the preset quality requirements are met.

[0009] As an optional approach, the raw material data in S1 above includes the purity, particle size, storage temperature and humidity of the insensitive energetic material, as well as the chip model, energy storage capacitor capacity deviation value, and ignition resistor resistance range of the detonator electronic control module. The production process data includes the welding temperature, assembly pressure, and transmission speed of the modified assembly line, as well as the low-voltage charging voltage, boost discharge duration, and dynamic adjustment coefficient of the energy feedback system. The test data includes ignition response time test values, impact sensitivity test results, thermal sensitivity test results, electrostatic sensitivity test results, radio frequency sensitivity test results, as well as safety level and national performance index breakdown data from third-party preliminary screening.

[0010] As an optional approach, in S2 above, before labeling outliers based on historical fault rework data, the historical fault rework data is further classified and organized, including: Obtain detailed information on rework losses for both ignition-free and ignition-charged electronic detonators, as well as production failure logs. Each log entry includes the time of failure, production stage, cause of failure, and direct losses. Historical failures are categorized into preset failure types based on core production processes, providing a classification basis for outlier weight labeling.

[0011] As an optional approach, the preset fault type classification includes: Slow charging voltage rise, discharge peak not reaching the threshold, or energy feedback system regulation failure are classified as abnormal discharge faults of electronic control module. The types of material failures, such as those with substandard purity, particle agglomeration, or excessive storage temperature and humidity, are classified as insensitive energetic material failures. The types of faults, such as welding temperature being too high or too low, assembly pressure deviation, or transmission speed being too high, are classified as process parameter deviation faults.

[0012] As an optional approach, in S2 above, the outlier weight annotation includes: S21. Remove invalid data without a clear cause of failure or loss record. Match valid data to preset failure types through keyword matching. If a single data point involves multiple types of failures, classify it according to the main cause of failure. S22. Using Python data processing tools, add weight scores to each abnormal data based on the degree of impact of economic losses and technical indicators of the fault, and label the associated technical indicators; S23. Extract a preset proportion of labeled data as a validation set. If the prediction accuracy deviation of a certain type of fault weight data exceeds the preset value, recalculate the loss amount and the degree of influence of the indicators for that type of fault, and adjust the weight score until the validation is passed.

[0013] As an optional approach, in the above S3, the input data dimensions for large model analysis specifically include dynamic parameters of the electronic control module, characteristic data of insensitive energetic materials, assembly process structural parameters, and historical sensitivity test data.

[0014] As an optional approach, S3 above includes at least one feature extraction model and at least four specific prediction models: The feature extraction model is used to structurally integrate the input dynamic parameters of the electronic control module, the characteristic data of the insensitive energetic material, the assembly process structural parameters and historical sensitivity test data, extract the static features and time-series features from the data, and output a fused feature vector containing the core information of sensitivity index prediction. The specialized prediction models are used to predict the impact sensitivity, thermal sensitivity, electrostatic sensitivity, and radio frequency sensitivity of the non-ignition electronic detonator based on the fused feature vectors output by the feature extraction model, and output the prediction results of the corresponding sensitivity index.

[0015] As an optional approach, specific prediction models include: The impact sensitivity prediction model takes assembly process structural parameters and the filling density of blunt-sensitive energetic material as input data to predict impact energy. The thermal sensitivity prediction model, whose input data are the purity of the insensitive energetic material and the peak discharge temperature of the electronic control module, is used to predict the heating temperature. The electrostatic sensitivity prediction model takes the insulation performance parameters of the electronic control module and the thickness of the antistatic coating on the casing as input data, and outputs the electrostatic voltage threshold. The RF sensitivity prediction model takes the insulation performance parameters of the electronic control module and the thickness of the antistatic coating on the casing as input data, and outputs the RF power threshold.

[0016] The beneficial effects of this invention are as follows: This invention provides a ignition-free electronic detonator and its quality prediction and control method. It not only provides a structure for the ignition-free detonator, but also provides complete support by collecting quality-related data from the entire production process. It combines historical fault and rework data to label outlier risk weights to improve model effectiveness. With the help of a specialized model, it accurately outputs sensitivity index prediction results and quality risk warnings. Then, it uses closed-loop control to adjust the process of risky batches, which can mitigate risks in advance, reduce waste, optimize processes, and ultimately improve the quality stability and production efficiency of ignition-free electronic detonators. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a front cross-sectional view of the ignition-free electronic detonator of the present invention; Figure 2 This is a side sectional view of the ignition-free electronic detonator of the present invention; Figure 3 This is a schematic flowchart of the method for predicting and controlling the quality of the ignition-free electronic detonator of the present invention. Detailed Implementation

[0018] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0019] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0020] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0021] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0022] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0023] Example Please see Figure 1 and Figure 2 This embodiment provides a detonator without a ignition charge, which achieves stable initiation without a ignition charge through the following structure, as implemented in this embodiment: The base detonator, as the main structure of the entire device, provides installation and housing space for other components. Its opening is fitted with a molded plug, which, when fitted with the opening of the base detonator, forms a sealed structure, effectively isolating external moisture, dust, and other impurities, protecting internal components from corrosion. Inside the molded plug are leads used to connect the electric ignition head to the external detonator, forming a channel for transmitting electrical signals from the detonator to the inside of the detonator. Wire clamps are used to secure the leads; proper placement of the wire clamps prevents loosening or displacement of the leads during transportation and use due to pulling, vibration, etc., ensuring the stability of electrical signal transmission.

[0024] The electronic control module is installed inside the base detonator to achieve precise control and safe detonation. The module's surface is equipped with spring contacts, whose primary function is to establish a stable connection with external circuitry. The elastic contact of these spring contacts ensures that the electrical signal or energy transmission between the electronic control module and related circuitry is not affected by poor contact.

[0025] As an optional structural design, the basic detonator also contains an initiating explosive and a main explosive. The initiating explosive is highly sensitive to energy and detonates first under the energy released by the electric igniter; the main explosive then explodes under the ignition energy of the initiating explosive, ultimately producing the energy required for blasting. In this optional design, the injection-molded plug and the electronic control module are welded together. This connection method enhances the structural stability between the two, preventing loosening under external environmental influences and further ensuring the reliability of the entire detonator's internal circuit connections and structural fit.

[0026] Please see Figure 3 This embodiment also provides a method for predicting and controlling the quality of ignition-free electronic detonators, including: S1. Collect raw material data, production process data and test data related to quality throughout the entire production process of ignition-free electronic detonators; S2. The collected data is cleaned and standardized. At the same time, outlier weights are labeled based on historical fault rework data to generate a training sample set with risk priority. S3. Construct a feature extraction model and a special prediction model, input the data preprocessed in S2 above, and output the sensitivity index prediction results and quality risk warning of the electronic detonator without ignition powder. S4. Based on the output of the large model, execute subsequent production processes for qualified batches, and adjust the process parameters for risky batches before re-collecting data and performing secondary analysis until the preset quality requirements are met.

[0027] In one implementation scenario, the above-mentioned solution aims to elaborate on the quality prediction and control method for ignitionless electronic detonators. This method is mainly used to solve the quality risk warning and production control problems in the production process of ignitionless electronic detonators, and to ensure that the products meet the safety performance, quality stability and cost control objectives proposed in the "Project Proposal for Ignitionless Electronic Detonators".

[0028] In this embodiment, S1 involves collecting quality-related data on raw materials, production processes, and testing throughout the entire production process of ignitionless electronic detonators. The reason for collecting these three types of data is that the quality of ignitionless electronic detonators is directly affected by the characteristics of raw materials, production process parameters, and finished product testing results. Only by comprehensively acquiring this data can complete and accurate basic information be provided for subsequent quality analysis and prediction. In actual production, data collection needs to cover the entire process from raw material warehousing to finished product testing to ensure no data omissions and lay a data foundation for subsequent steps.

[0029] In S1, the optional raw material data provided in this embodiment includes the purity, particle size, and storage temperature and humidity of the insensitive energetic material, as well as the chip model, energy storage capacitor capacity deviation, and ignition resistor resistance range of the detonator electronic control module. The insensitive energetic material is the core material for ignition in a priming-free electronic detonator. Its purity directly affects its ignition reliability; if the purity is substandard and contains impurities, ignition may be difficult. Particle size relates to energy accumulation; particle agglomeration leads to uneven energy distribution, affecting ignition efficiency. Excessive storage temperature and humidity reduce material stability, thus affecting product quality. The electronic control module is a key component for priming-free ignition. The chip model determines the module's control precision, the energy storage capacitor capacity deviation affects discharge energy, and the ignition resistor resistance range is related to instantaneous electrical energy generation. The acquisition of this data ensures quality control from the source of the raw materials.

[0030] Production process data includes welding temperature, assembly pressure, and transmission speed of the modified assembly line, as well as the low-voltage charging voltage, boost discharge duration, and dynamic adjustment coefficient of the energy feedback system. The modified assembly line needs to be adapted to the production process of ignition-free electronic detonators. Excessively high or low welding temperatures can lead to poor connection between the electronic control module and the tube shell, affecting product structural stability. Deviations in assembly pressure can cause uneven density of the insensitive energetic material, thus affecting energy accumulation. Excessively high transmission speeds may lead to gaps in process connections and assembly defects. The low-voltage charging voltage and boost discharge duration of the energy feedback system directly affect the discharge energy, while the dynamic adjustment coefficient relates to the adaptability of the discharge parameters. Collecting this data allows for real-time monitoring of key process parameters during production, preventing quality problems caused by process deviations.

[0031] The testing data includes ignition response time test values, impact sensitivity test results, thermal sensitivity test results, electrostatic sensitivity test results, radio frequency sensitivity test results, as well as the safety level and national performance index breakdown data from third-party preliminary screening. Ignition response time is a key performance indicator for ignition-free electronic detonators, requiring a minimum of 1 millisecond; its test value directly reflects the ignition response speed. Impact sensitivity, thermal sensitivity, electrostatic sensitivity, and radio frequency sensitivity test results are core indicators of product safety performance, directly related to safety during production, transportation, storage, and use. Optionally, the safety level from third-party preliminary screening must pass GB28263-2024 intrinsic safety certification, and the national performance index breakdown data must comply with WJ / T9085-2024 requirements. Collecting this testing data comprehensively assesses the product's performance and safety level, providing a basis for quality judgment.

[0032] S2 involves cleaning and standardizing the collected data, and weighting outliers based on historical fault rework data to generate a training sample set with risk priority. Data cleaning primarily removes outliers caused by equipment failures, such as welding data from instantaneous ultra-high temperatures on assembly lines. This type of data cannot accurately reflect normal production conditions, and retaining it would affect the accuracy of subsequent analysis. Standardization involves converting parameters with different units to the same range. For example, in production practice, the millisecond unit of discharge duration and the volt unit of charging voltage are usually normalized to the range of 0 to 1. This avoids biases in model analysis caused by unit differences. Weighting outliers based on historical fault rework data allows subsequent large models to prioritize learning outlier features that significantly impact quality, thereby improving the accuracy of quality prediction and ensuring the standardization of the generated training sample set.

[0033] This includes acquiring historical rework data on faults, such as detailed rework losses for both ignition-free and ignition-based electronic detonators, and a production fault log. The detailed rework loss details cover losses under different processes. The production fault log records specific faults in past production, with each entry including the fault occurrence time, production stage, cause, and direct loss. The fault occurrence time helps trace the production environment and process status at the time of the fault; the production stage identifies which phase the fault occurred in; the cause provides a basis for subsequent optimization; and the direct loss quantifies the impact of the fault.

[0034] This embodiment also categorizes historical faults into preset fault types based on production stages, providing a classification basis for outlier weight labeling. The production stages of a detonator without ignition charge include the operation of the electronic control module, the processing of insensitive energetic materials, and the implementation of assembly processes. Fault type classification must revolve around these stages. This classification method makes subsequent outlier weight labeling more targeted, allowing the model to clearly identify fault characteristics in different production stages during the learning process. This avoids model learning bias caused by mixed fault types, thereby improving the accuracy of subsequent quality predictions.

[0035] As an optional approach, the first type of fault provided in this embodiment is an abnormal discharge fault of the electronic control module that directly affects the ignition reliability of the detonator and may even lead to detonator failure. This fault is classified based on the abnormal behavior of the electronic control module during the discharge process, specifically including slow charging voltage rise, discharge peak not reaching the threshold, or energy feedback system adjustment failure. The electronic control module is the core control unit for ignition in a ignition-free electronic detonator. Slow charging voltage rise will prevent the energy storage capacitor from being fully charged in time; a discharge peak not reaching the threshold will prevent the generation of sufficient instantaneous electrical energy to ignite the insensitive energetic material at high temperature; and energy feedback system adjustment failure will prevent the discharge parameters from dynamically adapting to actual needs.

[0036] The second type of failure involves insensitive energetic materials that affect ignition performance. This is categorized based on abnormalities in the material's properties and storage conditions, specifically including substandard material purity, particle agglomeration, or excessive storage temperature and humidity. Substandard purity (containing impurities) reduces the material's ignition sensitivity, preventing proper ignition. Particle agglomeration causes uneven energy accumulation within the confined space, affecting energy delivery and hindering effective ignition of the initiating explosive. Excessive storage temperature and humidity alter the material's physicochemical properties, reducing its stability and impacting its ignition performance.

[0037] The third type of failure is a deviation in process parameters that affects the assembly quality and subsequent performance of the product. This is based on abnormal parameters during the assembly process, specifically including excessively high or low welding temperatures, assembly pressure deviations, or excessively high transmission speeds. Excessively high welding temperatures can damage the electronic control module or detonator housing, while excessively low temperatures can lead to weak connections, affecting the product's structural integrity. Assembly pressure deviations can cause uneven density of the insensitive energetic material, affecting energy accumulation. Excessively high transmission speeds can result in insufficient connection between processes, leading to assembly omissions.

[0038] The S2 described herein also provides outlier weight annotation. This annotation process is used to ensure data validity and weight accuracy, providing high-quality samples for subsequent model training. It includes: S21. Eliminate invalid data without a clear cause of failure or loss records. Use keyword matching to map valid data to preset failure types. If a single data point involves multiple failure types, classify it based on the primary cause. Invalid data without a clear cause of failure or loss records cannot provide effective information for the model; eliminating this data avoids interfering with model learning. Keyword matching determines the failure type by identifying key expressions in the data. In one optional scenario, keywords are set for types where the discharge peak does not reach the threshold to match the data to an abnormal discharge failure of the electronic control module; keywords are set for types where the material purity is substandard to match the data to a failure of insensitive energetic materials. When a single data point involves multiple failure types, such as an abnormal discharge of the electronic control module simultaneously causing the insensitive energetic material to fail to ignite, it must be classified based on the primary cause, i.e., classified as an abnormal discharge failure of the electronic control module. This ensures that each valid data point accurately corresponds to a preset failure type, laying the foundation for subsequent weight assignment.

[0039] S22. Using Python data processing tools, assign weight scores to each abnormal data point based on the degree of impact of economic losses and technical indicators, and label the associated technical indicators. The determination of weight scores needs to comprehensively consider two factors. One is the economic loss of the failure. Based on the calculation logic of the annual cost savings of electronic detonators without ignition powder in history, the loss of a single failure can be converted into a unit loss coefficient. The greater the loss, the higher the weight score. On the other hand, the impact of the fault on technical indicators is considered. The purpose is to label the associated technical indicators while adding weighted scores, allowing the subsequent model to clearly understand which core indicators the abnormal data affects. For example, labeling a data point as being associated with the fire response time indicator helps the model learn the abnormal characteristics related to that indicator. If the fault is directly associated with key technical indicators such as a fire response time of no more than 1 millisecond, passing GB28263-2024 intrinsic safety certification, and a fault rate of no more than 0.005%, the weighted score needs to be increased by 20% to 30%.

[0040] S23. Extract a preset proportion of labeled data as a validation set. If the prediction accuracy deviation of a certain type of fault weight data exceeds the preset value, recalculate the loss amount and the degree of impact on the indicators for that type of fault, and adjust the weight scores until the validation passes. The preset proportion is usually set to 30%. Extracting this part of the data as a validation set is to verify the rationality of the weight labeling. Prediction accuracy refers to the degree of matching between the weight data and the actual quality non-compliance results. The preset value is usually set to 10%. If the prediction accuracy deviation of a certain type of fault weight data exceeds 10%, for example, if the expected accuracy of a certain type of fault weight data is 75% but the actual accuracy is only 60%, then it is necessary to recalculate the loss amount of that type of fault, confirm whether there is a calculation error, and at the same time reassess its degree of impact on the technical indicators. Adjust the weight scores according to the assessment results, and perform validation again until the prediction accuracy deviation is within the preset value, ensuring that the weight labeling is accurate and reliable.

[0041] Subsequently, a large-scale model was constructed, comprising a feature extraction model and a specific prediction model. The input was the preprocessed data from S2, and the output was the predicted sensitivity index and quality risk warning for the ignition-free electronic detonator. The model's approach was to first integrate the preprocessed data using a feature extraction model to extract key features, and then use a specific prediction model to predict the safety indicators of the ignition-free electronic detonator. The feature extraction model was used to filter out crucial information affecting quality from complex data, while the specific prediction model was used for the sensitivity index, a key safety performance indicator.

[0042] The input data dimensions in S3 specifically include dynamic parameters of the electronic control module, characteristic data of insensitive energetic materials, assembly process structural parameters, and historical sensitivity test data.

[0043] Among them, the dynamic parameters of the electronic control module directly affect its working performance, including the discharge peak voltage, discharge duration, and energy regulation frequency of the energy feedback system. The discharge peak voltage and discharge duration determine whether the instantaneous electrical energy and high temperature released by the energy storage capacitor are sufficient to ignite the insensitive energetic material, while the energy regulation frequency affects the dynamic adaptability of the discharge parameters, ensuring stable discharge under different operating conditions.

[0044] The characteristic data of insensitive energetic materials determine their ignition performance and energy release effect, including material purity, particle size distribution, packing density, and micro-confined space volume. Material purity affects ignition sensitivity, particle size distribution affects energy accumulation uniformity, and packing density and micro-confined space volume jointly determine the energy accumulation effect. Inputting these data allows the model to understand the characteristic state of the insensitive energetic material and determine whether it can meet the ignition requirements.

[0045] Assembly process structural parameters directly affect the product's structural stability and resistance to external impacts, including the thickness of the detonator casing, the welding strength between the electronic control module and the casing, and the compaction degree of the insensitive energetic material filling.

[0046] Historical sensitivity test data can be set according to actual conditions. This embodiment may include raw test data for impact sensitivity, thermal sensitivity, electrostatic sensitivity, and radio frequency sensitivity, as well as sensitivity comparison data between traditional detonator and non-detonator. The raw test data provides a reference standard for the model's sensitivity index, while the comparative data allows the model to clearly understand the advantages and characteristics of non-detonator in sensitivity performance, helping the model to more accurately predict the sensitivity index of non-detonator electronic detonators.

[0047] In S3, at least one feature extraction model is used to structurally integrate the input dynamic parameters of the electronic control module, the characteristic data of insensitive energetic materials, the assembly process structural parameters, and historical sensitivity test data. During integration, different types of data are first standardized in format and matched in dimensions to ensure collaborative analysis. Then, static and temporal features are extracted from the data. Static features include parameters that do not change over time, such as material purity and shell thickness, while temporal features include parameters that change over time, such as discharge peak voltage and discharge duration. After feature extraction, key correlation features need to be strengthened, including the correlation between discharge parameters and energy accumulation in insensitive energetic materials, and the correlation between assembly structure and impact resistance. These correlation features have a significant impact on product quality. Finally, a fused feature vector containing sensitivity index prediction information is output, providing high-quality feature input for specialized prediction models.

[0048] In this embodiment, the feature extraction model may optionally employ a CNN-LSTM hybrid architecture, and its prediction-related data processing and feature extraction process includes: Input layer data processing: The dynamic parameters of the input electronic control module, the characteristic data of the insensitive energetic material, the assembly process structural parameters, and the historical sensitivity test data are structurally integrated. The dynamic parameters of the electronic control module are collected at preset time intervals to form time series data, and a time series feature matrix is ​​constructed based on this time series data. The characteristic data of the insensitive energetic material and the assembly process structural parameters are used as static data to construct a static feature matrix. After dimensionality matching between the time series feature matrix and the static feature matrix, they are converted into 128-dimensional feature vectors, which are used as input data for the main model.

[0049] The feature fusion layer processing includes static feature extraction, temporal feature extraction, and attention mechanism enhancement. Static feature extraction uses CNN to extract features from the static feature matrix, with a 3×3 convolution kernel and a stride of 1. Convolution operations are performed on static features such as purity of insensitive energetic materials, particle size distribution, filling density, volume of small closed spaces, thickness of detonator shell, welding strength between electronic control module and shell, and compaction degree of insensitive energetic material filling to generate static feature maps. Then, max pooling is used to downsample the static feature maps to retain key static feature information.

[0050] The time series feature extraction uses an LSTM network to extract features from the time series feature matrix. It sets up 64 hidden layer units and uses a gating mechanism to capture the dynamic change features of time series data such as the peak discharge voltage, discharge duration, and energy regulation frequency of the electronic control module. It learns the correlation between data at different time nodes and outputs a time series feature vector.

[0051] An attention mechanism is then introduced to fuse static and temporal features, and the attention weight of each feature is calculated. This weight is determined based on the degree of influence of the feature on the sensitivity index of the ignition-free electronic detonator. Specifically, the attention weight of the features related to the discharge parameters and the energy accumulation of the insensitive energetic material, and the features related to the assembly structure and the impact resistance are increased to enhance the related features. The enhanced static features and temporal features are then concatenated to obtain the fused feature matrix.

[0052] The output layer flattens the fused feature matrix and maps it into a 64-dimensional fused feature vector through a fully connected layer. This vector contains the feature information required for predicting the sensitivity index of the ignition-free electronic detonator and is directly input into the special prediction model.

[0053] This embodiment employs at least four specialized prediction models, corresponding to four safety indicators of the ignition-free electronic detonator: impact sensitivity, thermal sensitivity, electrostatic sensitivity, and radio frequency sensitivity. Each specialized prediction model performs predictions based on the fused feature vector output by the feature extraction model. Since the influencing factors and evaluation criteria for the four sensitivity indicators differ, each specialized prediction model focuses specifically on the feature information corresponding to itself. For example, the impact sensitivity prediction model focuses on analyzing features related to impact resistance, while the thermal sensitivity prediction model focuses on analyzing features related to thermal stability. Through this design, each specialized prediction model can more accurately capture the influencing factors of its corresponding sensitivity indicator, thereby outputting accurate sensitivity indicator prediction results and ensuring comprehensive coverage of the safety performance prediction requirements for ignition-free electronic detonators.

[0054] This embodiment sets up four specific prediction models, all of which use gradient boosting regression models to make predictions in the aforementioned directions. The prediction process of each sub-model is set up as follows: For the impact sensitivity prediction sub-model, the features related to assembly process structural parameters and the density of insensitive energetic material filling are taken from the 64-dimensional fused feature vector output by the main feature extraction model as input features. Impact test data of ignitionless electronic detonators obtained during the technology verification phase are used as training labels. These labels represent the measured impact energy corresponding to a 50% ignition probability of the ignitionless electronic detonator under different test conditions.

[0055] A gradient boosting regression model is used to fit the input features to the training labels, with the mean absolute error (MAE) as the loss function. The loss function is calculated using the following formula: in n The number of training samples. y i For the first i Measured impact energy values ​​for each sample. ŷ i For the first i The model outputs the predicted impact energy of each sample; by iteratively optimizing and minimizing the loss function, the model outputs the predicted impact energy corresponding to a 50% ignition probability of an electronic detonator without ignition powder.

[0056] For the thermal sensitivity prediction model, the features related to the purity of the insensitive energetic material and the peak discharge temperature of the electronic control module from the 64-dimensional fused feature vector output by the main feature extraction model are used as input features. Thermal sensitivity test data of the ignition-free electronic detonator obtained during the technology verification phase are used as training labels. These labels represent the measured heating temperatures corresponding to a 50% ignition probability for the ignition-free electronic detonator under different test conditions.

[0057] The input features and training labels are fitted and trained using a gradient boosting regression model, with the mean absolute error used as the loss function. The loss function is calculated using the same formula as the impact sensitivity prediction sub-model. The loss function is minimized through iterative optimization, so that the model outputs the predicted heating temperature corresponding to a 50% ignition probability for an electronic detonator without ignition powder.

[0058] For the electrostatic sensitivity prediction model, the features related to the insulation performance parameters of the electronic control module and the thickness of the antistatic coating on the tube shell from the 64-dimensional fused feature vector output by the main feature extraction model are used as input features. Electrostatic sensitivity test data of the ignition-free electronic detonator obtained during the technology verification phase are used as training labels. These labels represent the measured electrostatic voltage values ​​corresponding to a 50% ignition probability for the ignition-free electronic detonator under different test conditions.

[0059] The input features and training labels are fitted and trained using a gradient boosting regression model, with the mean absolute error used as the loss function. The loss function is calculated using the same formula as the impact sensitivity prediction sub-model. The loss function is minimized through iterative optimization, so that the model outputs the predicted electrostatic voltage threshold value corresponding to a 50% ignition probability of an electronic detonator without ignition powder.

[0060] For the RF sensitivity prediction sub-model, the features related to the insulation performance parameters of the electronic control module and the thickness of the antistatic coating on the casing are taken from the 64-dimensional fused feature vector output by the feature extraction main model as input features.

[0061] The radio frequency sensitivity test data of the ignitionless electronic detonator obtained during the technology verification phase was used as the training label. The label is the measured value of the radio frequency power corresponding to the 50% ignition probability of the ignitionless electronic detonator under different test conditions.

[0062] The input features and training labels are fitted and trained using a gradient boosting regression model, with the mean absolute error used as the loss function. The loss function is calculated using the same formula as the impact sensitivity prediction sub-model. The loss function is minimized through iterative optimization, so that the model outputs the predicted value of the radio frequency power threshold corresponding to a 50% ignition probability of an electronic detonator without ignition powder.

[0063] Finally, based on the output of the large-scale model, subsequent production processes are executed for qualified batches, while for risky batches, process parameters are adjusted, data is re-collected, and a second analysis is performed until the preset quality requirements are met. This step is a crucial link in applying the model analysis results to actual production control. For batches determined to be qualified by the large-scale model, they can directly proceed to the next production process; for batches determined to be risky, the current production process is paused, and optimization is performed based on the process parameter adjustment suggestions output by the model. After optimization, relevant data is re-collected and input into the model for a second analysis. Only after the second analysis is qualified and the preset quality requirements are met can the next process proceed, ensuring that the final product quality meets the standards.

[0064] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An electronic detonator without a primer, characterized in that, The application relates to a safety fuse, which comprises a line card, an injection plug, an electronic control module and a basic detonator. The injection plug is arranged at an opening of the basic detonator and is internally provided with a foot wire for connecting an electric ignition head and a detonator. The electronic control module is arranged in the basic detonator and is provided with a spring sheet on the surface for communicating with an external circuit. The basic detonator is internally provided with a primer and a main charge, and the injection plug and the electronic control module are welded.

2. The electronic detonator without priming powder according to claim 1, characterized in that, The application relates to a quality risk prediction method for a non-ignition electronic detonator.

3. A method for predicting and controlling the quality of an electronic detonator without a primer, characterized in that, S1, collecting raw material data, production process data and detection data related to quality in the whole production process of the non-ignition electronic detonator; S2, performing cleaning and standardization processing on the collected data, and performing abnormal value weight marking based on historical fault rework data to generate a training sample set with risk priority; S3, constructing a feature extraction model and a special prediction model, inputting the data preprocessed in S2, and outputting a sensitivity index prediction result and a quality risk early warning of the non-ignition electronic detonator; S4, according to the output result of the large model, executing subsequent production procedures on qualified batches, adjusting process parameters on risk batches, re-collecting data and performing secondary analysis until the preset quality requirement is met. In S1, the raw material data comprises the component purity, granularity, storage temperature and humidity of the insensitive energetic material, the chip model of the electronic control module of the detonator, the capacity deviation value of the energy storage capacitor and the resistance value range of the ignition resistor; 4. The quality prediction and control method of the non-priming electronic detonator according to claim 3, characterized in that, The production process data comprises the welding temperature, assembly pressure and transmission speed of the reformed assembly line, and the low-voltage charging voltage, voltage boosting and discharging time length and dynamic adjustment coefficient of the energy feedback system; The detection data comprises a firing response time test value, impact sensitivity test result, thermal sensitivity test result, electrostatic sensitivity test result, radio frequency sensitivity test result, safety level of third-party preliminary screening and national performance index subitem data. In S2, before the abnormal value weight marking based on the historical fault rework data, historical fault rework data classification and analysis are further included, which comprises:

5. The quality prediction and control method of the non-priming electronic detonator according to claim 3, characterized in that, Obtaining fault rework loss details and production fault account books of the non-ignition electronic detonator and the ignition electronic detonator, and each account data obtained comprises fault occurrence time, production link, fault reason and direct loss; Dividing the historical faults into preset fault types according to production links to provide classification basis for abnormal value weight marking. The division of the preset fault types comprises:

6. The non-priming electronic detonator quality prediction and control method according to claim 5, characterized in that, Dividing the type of slow charging voltage climbing, discharging peak value not reaching a threshold value or energy feedback system adjustment failure into an electronic control module discharging abnormal fault; Dividing the type of material purity not reaching a standard, particle agglomeration or storage temperature and humidity exceeding a standard into an insensitive energetic material fault; Dividing the type of welding temperature being too high or too low, assembly pressure deviation or transmission speed being too fast into a process parameter deviation fault. In S2, the abnormal value weight marking comprises:

7. The quality prediction and control method of the non-priming electronic detonator according to claim 5, characterized in that, S21, removing invalid data without an explicit fault reason and loss record, matching the effective data to the preset fault types through a keyword, and classifying according to a main fault reason if a single data involves multiple fault types; ​ S22, using Python data processing tools, based on the economic loss of the fault and the influence degree of the technical index, adding weight score to each abnormal data, and marking the associated technical index; S23, extract a preset proportion of labeled data as a validation set, if the prediction accuracy of the weight data of any type of fault deviates more than a preset value, recalculate the loss amount and index influence degree of the fault, adjust the weight score until the verification is passed.

8. The quality prediction and control method of the non-priming electronic detonator according to claim 3, characterized in that, In the above S3, the input data dimension of the large model analysis specifically includes electronic control module dynamic parameters, insensitive energetic material characteristic data, assembly process structure parameters and historical sensitivity test data.

9. The quality prediction and control method of the non-priming electronic detonator according to claim 3, characterized in that, In the above S3, at least one feature extraction model and at least four special prediction models are included: The feature extraction model is used to structure and integrate the input electronic control module dynamic parameters, insensitive energetic material characteristic data, assembly process structure parameters and historical sensitivity test data, extract static features and time sequence features in the data, and output a fusion feature vector containing sensitivity index prediction core information; The special prediction model is respectively used to predict the impact sensitivity, thermal sensitivity, electrostatic sensitivity and radio frequency sensitivity of the non-priming electronic detonator based on the fusion feature vector output by the feature extraction model, and output the prediction results of the corresponding sensitivity index.

10. The non-priming electronic detonator quality prediction and control method according to claim 9, characterized in that, The special prediction model includes: The impact sensitivity prediction model, whose input data is the assembly process structure parameter and the filling density of insensitive energetic material, is used to predict the impact energy; The thermal sensitivity prediction model, whose input data is the purity of insensitive energetic material and the discharge peak temperature of electronic control module, is used to predict the heating temperature; The electrostatic sensitivity prediction model, whose input data is the insulation performance parameter of electronic control module and the thickness of pipe shell anti-static coating, is used to output the electrostatic voltage threshold; The radio frequency sensitivity prediction model, whose input data is the insulation performance parameter of electronic control module and the thickness of pipe shell anti-static coating, is used to output the radio frequency power threshold.

Images