Electronic detonator full cycle data management method and system

Abstract

The application provides an electronic detonator full-cycle data management method and system, relates to the technical field of data management, and comprises the following steps: creating an append-only non-modified life cycle data chain for each electronic detonator, recording stage data and uncertainty measurement; constructing an initial normal probability distribution characterized by mean and variance based on a multi-dimensional quality characteristic vector and a measurement standard deviation in a production stage; continuously evolving the probability distribution based on an attenuation model and environmental data; compressing cognitive uncertainty through Bayesian updating, and realizing information sharing and divergence detection of detonators in the same batch through a batch association mechanism modulated by information purity; calculating reliability based on a posterior probability distribution, and performing risk quantization distribution matching; triggering attenuation model parameter calibration through use stage actual measurement feedback, and reversely updating the state of all in-stock detonators, so that full-cycle closed-loop management from uncertainty modeling to decision utilization is realized.

Description

Technical Field

[0001] This invention relates to the field of data management technology, specifically to a method and system for full-cycle data management of electronic detonators. Background Technology

[0002] Electronic detonators are widely used in civil blasting projects such as mining, tunneling, and urban demolition. As a core pyrotechnic product for achieving precise delayed detonation, with the deepening of digital transformation in the civil explosives industry, the production and use of electronic detonators have continued to grow, which has placed higher demands on their full-cycle quality control capabilities from production and warehousing to on-site use. In the existing technology, the production management, storage management and detonation control of electronic detonators are handled by independent information systems. The systems are linked by identification codes. Storage management mainly adopts a distribution strategy with fixed shelf life and first-in-first-out. Quality status assessment relies on the conclusion of a single test at the time of leaving the factory as the basis for judging the entire life cycle. The existing technology has the following shortcomings: data at each stage are independent of each other, forming information silos; the quality status lacks a continuous update and dynamic evaluation mechanism throughout the entire life cycle; and the distribution decision is based solely on the order of entry into the warehouse without considering the actual quality status of each detonator and its suitability for the usage scenario, resulting in limited inventory turnover efficiency and insufficient refined management capabilities. Summary of the Invention

[0003] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a full-cycle data management method and system for electronic detonators. This addresses the problems mentioned in the background technology, such as data at each stage being independent and forming information silos, lack of continuous updates and dynamic evaluation mechanisms for quality status throughout the entire lifecycle, and distribution decisions being based solely on the order of entry into the warehouse without considering the actual quality status of each detonator and its suitability for the usage scenario, resulting in limited inventory turnover efficiency and insufficient refined management capabilities.

[0004] Technical solution To achieve the above objectives, the present invention is implemented through the following technical solution: a method and system for full-cycle data management of electronic detonators, comprising, S1, creating a block-based lifecycle data chain for each electronic detonator that is only appended and not modified, including a production block, a warehousing block, an environment block, a sampling block, a distribution block, and a usage feedback block, wherein each block records the corresponding stage data and uncertainty measurement; S2, construct an initial normal probability distribution based on the multi-dimensional quality feature vector and measurement standard deviation collected by the coding detection terminal, and represent it with the initial mean vector and the initial variance diagonal matrix, and write it into the production block along with the production batch number and the production completion timestamp; S3, based on the temperature and humidity time series data, perform state transition operation with the currently effective decay model, subtract the mass decay increment vector from the mean vector of the normal probability distribution to obtain the transition mean vector, add the variance diagonal matrix to the process noise covariance matrix to obtain the transition variance diagonal matrix, and write it into the environmental block; S4, Receive the observation vector of quality sampling inspection, update the observation vector with the transition mean vector and transition variance diagonal matrix in the environment block based on Bayesian update, obtain the posterior mean vector and posterior variance diagonal matrix, and write them into the sampling inspection block; S5, receive an engineering distribution request that includes a precision requirement threshold and an acceptable maximum risk level, and calculate the reliability of the stocked electronic detonators in an allocable state based on the posterior mean vector and the posterior variance diagonal matrix. The reliability is calculated by subtracting the difference of the delay time precision dimension components in the posterior mean vector from the precision requirement threshold, dividing by the square root of the variance of the delay time precision dimension in the posterior variance diagonal matrix, and then substituting it into the standard normal cumulative distribution function. S6. Arrange the electronic detonators in the allocable state according to their reliability, select the required number of electronic detonators from the top engineering distribution requests to form a candidate allocation list, check whether the minimum reliability value meets the acceptable maximum risk level constraint, if it does, confirm the allocation scheme and write it into the distribution block.

[0005] Preferably, the electronic detonator is referred to as a detonator below. The system consists of six functional modules: a lifecycle data chain storage module, a Bayesian state estimation engine, a batch association update module, an information value-driven verification and scheduling module, a risk quantification distribution and matching module, and an edge data adapter. The edge data adapter is deployed at the boundary of each stage management subsystem and is responsible for converting the raw data format into a standardized data block format and writing it into the lifecycle data chain storage module. The six functional modules are connected sequentially in the data flow direction to form a closed-loop structure. The output of the risk quantification distribution and matching module is fed back to the Bayesian state estimation engine through the detonation control system to achieve model parameter calibration. A unique lifecycle data chain is created for each detonator from the time of production coding, using an append-only, non-modifying approach. The block-based structure includes six types: production block, warehousing block, environmental block, sampling block, distribution block, and usage feedback block. Each block records data and uncertainty measures for the corresponding stage. Probabilistic modeling is achieved by transforming the quality status from a deterministic numerical value to a normal probability distribution characterized by mean and variance. The mean represents the optimal quality estimate based on known information, and the variance represents the degree of cognitive uncertainty. The normal probability distribution evolves sequentially in four stages: an initial probability distribution is established through the production stage; mean degradation and variance expansion are driven by environmental data in the storage stage; variance compression is driven by observation information in the sampling stage; and model parameter calibration is triggered by stage-based measured feedback, and the probability distribution of all detonators in the warehouse is updated in reverse through a batch association mechanism, forming a complete management closed loop.

[0006] Preferably, during the production stage, after a detonator completes coding and quality inspection through a coding and detection terminal, the terminal acquires a multi-dimensional quality feature vector. This multi-dimensional quality feature vector includes measured values ​​for four dimensions: assembly deviation, curing degree, energy storage capacitor capacity, and clock oscillator frequency. Simultaneously, the accuracy level of the measuring equipment for each dimension is recorded. Based on the accuracy level, the standard deviation of each dimension is determined. Specifically, the standard deviation of assembly deviation is 0.005 mm, the standard deviation of curing degree is 0.02%, the standard deviation of energy storage capacitor capacity is 0.5 μF, and the standard deviation of clock oscillator frequency is 0.1 kHz. An initial normal probability distribution, denoted as Q0, is constructed based on the multi-dimensional quality feature vector and the standard deviations of each dimension, following a normal distribution with mean μ0 and variance Σ0. The initial mean vector μ0... The value is a 4-dimensional vector composed of the measurement values ​​P1 to P4 of each dimension, where P1 is the assembly deviation measurement value, P2 is the curing degree measurement value, P3 is the energy storage capacitor capacity measurement value, and P4 is the clock oscillator frequency measurement value. The initial variance diagonal matrix Σ0 is set as a 4x4 diagonal matrix. The diagonal elements of the variance diagonal matrix are the squares of the standard deviations of each dimension measurement, and the off-diagonal elements are all set to zero, indicating that the four dimensions are measured independently and there is no correlation between the dimensions. Then, the mean vector and the variance diagonal matrix are written into the production block of the corresponding detonator in the life cycle data chain storage module. At the same time, the unique identification, production batch number and production completion timestamp of the detonator are recorded. Through the above construction process, the transformation from deterministic measurement values ​​to probabilistic cognitive expression is completed, providing the initial input for the Bayesian state estimation engine to perform state transition operations.

[0007] Preferably, after the detonators are received into the warehouse, the Bayesian state estimation engine begins to perform state transition operations. Temperature and humidity time-series data for the warehouse containing the detonators are continuously collected hourly by temperature and humidity sensors deployed in various storage locations. The temperature sampling accuracy is ±0.5℃, and the humidity sampling accuracy is ±2%. The Bayesian state estimation engine uses the currently active attenuation model as its computational basis, starting from the initial normal probability distribution recorded in the production block. Each time a new set of temperature and humidity sampling data is received, a state transition operation is performed. The currently active attenuation model in the initial stage... A preset attenuation model is adopted. After subsequent model parameter calibration, the model is updated to the calibrated attenuation model. The currently effective attenuation model substitutes the temperature sampling data into the Arrhenius acceleration factor function to calculate the temperature acceleration factor. The Arrhenius acceleration factor function is calculated using an exponential function with the natural constant e as the base. The exponent is equal to the ratio of the activation energy parameter to the universal gas constant multiplied by the reciprocal of the reference temperature minus the reciprocal of the current measured temperature. The activation energy parameter Ea is in eV, and the preset initial value is 1.0 eV. The universal gas constant R, i.e., the Boltzmann constant, is approximately 8.314 J·mol. -1 ·K -1 The initial value of the reference temperature T0 is set to 293K, which is equivalent to a standard storage temperature of 20℃. The humidity diffusion coefficient is calculated by substituting the humidity sampling data into the diffusion coefficient function D(H). The expression is The fundamental diffusion coefficient constant The preset initial value is 0.001 / h, and the humidity sensitivity coefficient is... The preset initial value is 0.02. The diffusion coefficient function is calculated by multiplying the basic diffusion coefficient constant by the humidity acceleration factor, specifically: ,in: The value represents the current measured relative humidity as a percentage. The humidity acceleration factor is: The product of the temperature acceleration factor and the humidity diffusion coefficient is determined as the mass decay increment vector for the current time step. This mass decay increment vector is a 4-dimensional vector, with each dimension corresponding to the microstructure creep increment in the assembly deviation direction, the colloidal aging increment in the curing degree direction, the electrochemical degradation increment in the energy storage capacitor capacity direction, and the crystal drift increment in the clock oscillator frequency direction, respectively. The mass decay increment vector is subtracted from the probability distribution mean vector of the previous time step to obtain the transition mean vector for the current time step. The variance diagonal matrix of the previous time step is added to a preset process noise covariance matrix to obtain the transition variance diagonal matrix for the current time step. This process noise covariance matrix is ​​a 4x4 diagonal matrix, with diagonal elements taking values ​​of 0.0001 mm. 2 0.001% 2 0.1μF2 0.02kHz 2 If a sampling observation update or batch association update occurs between two state transition operations, the next state transition operation will start with the updated posterior mean vector and posterior variance diagonal matrix. After the state transition operation is completed, the temperature and humidity sampling values, temperature acceleration factor, humidity diffusion coefficient, transition mean vector, and transition variance diagonal matrix will be written into the environmental block of the corresponding detonator in the lifecycle data chain storage module.

[0008] Preferably, when a quality management instruction specifies a quality sampling inspection of a certain detonator in stock, the measured delay time of the detonator is obtained using a delay accuracy measuring device, and the measured energy storage capacitor capacity of the detonator is obtained simultaneously using a capacitance measuring device. These two measured values ​​together form an observation vector. The measurement error covariance matrix of the observation vector is determined according to the accuracy level of the measuring device. The standard deviation of the delay time measurement is 0.5 μs, and the standard deviation of the energy storage capacitor capacity measurement is 0.3 μF. The measurement error covariance matrix is ​​a 2x2 diagonal matrix, with the diagonal elements representing the delay time, delay time, and energy storage capacitor capacity respectively. The squares of the standard deviations of the time interval measurements and the standard deviations of the energy storage capacitor capacity measurements are taken, with off-diagonal elements set to zero. After receiving the observation vector, the Bayesian state estimation engine reads the transition mean vector and transition variance diagonal matrix of the current time step from the environmental block corresponding to the detonator in the lifecycle data chain storage module as prior distribution parameters. It maps the two components of the observation vector to the corresponding delay time precision dimension and energy storage capacitor capacity dimension in the 4-dimensional state space, constructing a 2x4 observation matrix with 1s in the corresponding dimension positions and 0s in the rest. The Kalman gain matrix is ​​then calculated using the Bayesian update formula. , The posterior mean is The posterior covariance is The posterior distribution is ,in, Let be the mean vector of the prior states, which are predictions of the current system state based on information from the previous time step. Let be the covariance matrix of the prior states; The observation matrix is ​​the state that is mapped to the observation space. The measurement error covariance matrix; This represents the observation vector at the current moment; This is the updated posterior mean vector; This is the updated posterior covariance matrix; The identity matrix is ​​used; the diagonal elements of the posterior variance diagonal matrix are all reduced compared to the prior variance diagonal matrix. After completing the Bayesian update, the observation vector, the measurement error covariance matrix, the posterior mean vector, and the posterior variance diagonal matrix are written into the sampling block of the corresponding detonator in the lifecycle data chain storage module. At the same time, the detonator is marked as a sampled detonator and its identity is recorded to provide an information allocation benchmark for subsequent batch association updates.

[0009] Preferably, detonators with the same batch number share the same production process parameters and raw material batches, resulting in a natural correlation in their quality status at the time of manufacture. Simultaneously, detonators stored in the same or adjacent storage locations also exhibit a correlation in their degradation trajectories during storage due to experiencing almost identical temperature and humidity environments. The batch correlation coefficient is determined according to the following four-layer rule: The first layer sets the basic correlation coefficient: a basic correlation coefficient of 0.5 is set between all detonators with the same production batch number, and a basic correlation coefficient of 0 is set between detonators with different production batch numbers. The second layer is storage location correlation correction: if the same... If two detonators from the same batch are currently stored in the same storage location, the storage location correlation increment is increased by 0.2. If they are stored in adjacent storage locations and the temperature and humidity difference does not exceed the preset temperature difference threshold of 2℃ and the preset humidity difference threshold of 5%, the storage location correlation increment is increased by 0.1. If the temperature and humidity difference exceeds the thresholds, the storage location correlation increment is 0. The third layer is transportation correlation correction: if two detonators from the same batch experienced the same transportation batch number before entering the warehouse and the vibration record does not exceed the preset vibration threshold of 0.5g, the transportation correlation increment is increased by 0.15. If the transportation batches are different or the vibration exceeds the standard, the transportation correlation increment is 0. The fourth layer is time decay correction. ,in, The batch correlation coefficient represents the final correlation strength between batch i and batch j. After calculation, it needs to be limited to the interval [0, 0.85] by a truncation function. The table shows the basic correlation coefficient, with a preset initial basic correlation value of 0.5, which represents the baseline correlation level when there are no additional correlation factors. The incremental value for warehouse location association is the association value increased due to the same or similar warehouse locations. The value is discrete: 0 for no association, 0.1 for general association, or 0.2 for strong association. The incremental value for transportation association is the association value increased due to shared transportation batches or routes. The value is discrete: 0 for no association or 0.15 for association. t is the time elapsed from the time of completion of joint production to the current calculation time, in days. The batch association coefficients are stored in the batch association update module in a sparse matrix format. Only detonator pairs with an association coefficient greater than the preset storage threshold of 0.05 are stored for subsequent batch association update calls.

[0010] Preferably, after the Bayesian observation update of a single detonator is completed, the batch correlation update module immediately performs a reverse confidence propagation update based on information purity. First, it calculates the information purity index of the sampled detonators. This index quantifies the consistency between the current observation information and the prior prediction. It is calculated by subtracting the sum of the diagonal elements of the posterior variance diagonal matrix after sampling from 1, compared to the sum of the diagonal elements of the prior variance diagonal matrix before sampling. The information purity index ranges from 0 to 1; a value close to 0 indicates a high degree of consistency between the observed data and the prior prediction, while a value close to 1 indicates a significant discrepancy. Then, it calculates the confidence propagation temperature coefficient, nonlinearly modulating the batch correlation coefficient and the information purity index. The modulation rule is the difference between the batch correlation coefficient multiplied by 1 and the square of the information purity index. When the information purity index is 0.5, the propagation strength retains 75% of the original correlation coefficient. When the information purity index is 0.8, the propagation intensity retains only 36% of the original correlation coefficient. The correlation coefficient between the sampled detonator and all detonators in the same batch is read from the sparse matrix of the batch correlation update module. Detonator pairs with a confidence propagation temperature coefficient greater than the preset update threshold of 0.1 are selected to form the affected detonator set for this update. For each detonator in the affected detonator set, a correlation update is performed. The information quantity is multiplied by the confidence propagation temperature coefficient, then multiplied by 1, minus the information purity index, and then divided by the sum of the number of affected detonators plus 1 to obtain the information allocation. The probability distribution mean vector of each detonator in the inventory is shifted towards the observation update direction of the sampled detonator. The corresponding component of the information allocation is subtracted from each diagonal element of the variance diagonal matrix, but the variance compression amplitude must not be lower than the preset minimum variance lower limit. The diagonal elements of the preset minimum variance lower limit are 0.000025mm. 2 0.0001% 2 0.04μF 2 0.0025kHz 2 When the information purity index exceeds the preset divergence threshold of 0.6, information propagation is not performed. Each sample in the divergence buffer queue contains eight fields: the unique identifier of the detonator being sampled, the transition mean vector at the sampling time, the transition variance diagonal matrix, the observation vector, the measurement error covariance matrix, the posterior mean vector, the posterior variance diagonal matrix, and the information purity index. This observation is marked as a suspicious observation and stored in the divergence buffer queue.

[0011] Preferably, when the number of samples in the divergence buffer queue reaches a preset cumulative number of 10, a fast calibration process for the currently effective decay model parameters is triggered, defining the parameter vector to be optimized. ,in The activation energy parameter, Given a constant diffusion coefficient, construct a weighted loss function: ,in Let θ represent the square of the Euclidean norm, which is the sum of the squares of the vector components. Then, gradient descent is used to iteratively update θ. Number of iterations The upper limit is set to 50 iterations, and the stopping condition is reaching the iteration limit or the gradient norm is small enough. The updated θ is the calibrated model parameters. Where N is the current number of samples in the divergence buffer queue, the trigger threshold is fixed at 10, and i is the sample index. All samples in the buffer queue are traversed. is the information purity index of the i-th sample, which serves as the weight of the sample in the loss function. The larger the value, the greater the influence of the sample on the parameter update. Let θ be the mean of the state predictions made for the i-th branch point in the buffer queue before the actual measured value of that point is seen; Let be the observation vector corresponding to the i-th sample; H is the observation matrix, which maps the state space to the observation space; θ is the parameter vector to be calibrated, which contains only two components: the activation energy parameter and the diffusion coefficient constant. Let θ be the gradient vector of the loss function. The learning rate is set to 0.01, and the maximum number of iterations is set to 50. After calibration, the divergence buffer queue is cleared, and a reverse re-evaluation is triggered. The Bayesian state estimation engine uses the updated decay model parameters to re-execute the complete state transition sequence from the time of entry into the warehouse for all detonators in the warehouse. For each detonator in the warehouse, the gradient of its probability distribution with respect to the decay model parameters is calculated. Only detonators whose absolute gradient value exceeds the preset sensitivity threshold of 0.01 are fully recalculated. For detonators whose absolute gradient value is lower than the threshold, a first-order Taylor expansion approximate update is used to reduce the computational load. The specific calculation method of the first-order Taylor expansion approximate update is as follows: taking the current decay model parameter θ0 as the expansion point, the first-order partial derivative of the state transition mean function with respect to θ is multiplied by the parameter update amount Δθ to obtain the approximate updated transition mean vector. That is, the update increment of the transition mean vector is equal to the Jacobian. Multiplying the matrix by the parameter update amount Δθ eliminates the need to re-execute the complete state transition sequence. After recalculation, the current posterior mean vector and posterior variance diagonal matrix of each detonator are updated. The risk quantification distribution and matching module recalculates the reliability of all detonators in the library based on the updated probability distribution and updates the risk labels. If the reliability of a batch of detonators drops from above 97% to below 97% due to model correction, a batch-level early warning notification is automatically generated. At the same time, the timestamp of the calibration event, the trigger path identifier, the information purity summary value, and the updated decay model parameters are written into the usage feedback block of the lifecycle data chain storage module of all affected detonators. The updated decay model parameters are written back to the Bayesian state estimation engine through the edge data adapter to replace the activation energy parameters and diffusion coefficient parameters in the previously effective decay model. All subsequent state transition operations of all detonators in the library are performed based on the updated decay model parameters.

[0012] Preferably, the information value-driven verification and scheduling module calculates an information value score for each detonator in the inventory. This information value score is determined by a combination of an uncertainty reduction potential index and a decision influence weight index. The uncertainty reduction potential index is calculated as follows: The variance diagonal matrix of the current probability distribution of the detonator is read; the sum of the four diagonal elements is obtained to get the total variance value; this is then divided by the sum of preset baseline variances to obtain a standardized uncertainty index. The preset baseline variance sum is equal to the sum of the squares of the standard deviations of each dimension, which is 0.260425. The larger the standardized uncertainty index, the more severe the cognitive uncertainty. The decision influence weight index is calculated as follows: The accuracy requirement threshold recorded in the most recent distribution request of the detonator is obtained; the absolute value of the difference between the component of the delay time accuracy dimension in the posterior mean vector and the accuracy requirement threshold is calculated; this is then divided by 10% of the accuracy requirement threshold to obtain a threshold sensitivity index. If the detonator has never participated in any engineering distribution request... Using a preset baseline accuracy threshold of 5μs as the default accuracy requirement threshold, a threshold sensitivity index is calculated. The standardized uncertainty index is multiplied by the threshold sensitivity index, and then multiplied by a preset comprehensive scaling factor of 100 to obtain the information value score, which ranges from 0 to 100. If the component of the delay time accuracy dimension in the posterior mean vector is lower than the preset minimum quality threshold of 50μs, the detonator is marked as prohibited from firing. The information value score of detonators in the prohibited state is forcibly set to zero. Detonators with information value scores greater than the preset queuing threshold of 10 are arranged in descending order of score to form a priority sampling queue. The maximum length of the priority sampling queue is set to the available sampling time of 8 hours divided by the average single firing time of 15 minutes, i.e., 32 firings. After each sampling is completed, the information value score of the affected detonators is recalculated, and the priority sampling queue is dynamically updated. Detonators with scores lower than 10 are removed, and new detonators exceeding the threshold are inserted into the corresponding positions. If the queue is empty, the sampling task of this management cycle ends early.

[0013] Preferably, after receiving an engineering distribution request, the risk quantification distribution matching module performs reliability calculation and risk quantification matching. The engineering distribution request includes three input parameters: the required number of detonators, the minimum accuracy level threshold, and the maximum acceptable risk level. Specifically, the delay time accuracy threshold for high-precision demolition blasting projects is 2 μs, and the maximum acceptable risk level is 1%. The delay time accuracy threshold for standard earthwork projects is 5 μs, and the maximum acceptable risk level is 5%. The risk quantification distribution matching module iterates through all projects in an assignable state. The detonators in the inventory that are in an allocatable state are those that are not marked as prohibited from firing and are not marked as red alert level. For each detonator in an allocatable state, the current posterior mean vector and posterior variance diagonal matrix are read from the latest block of its lifecycle data link storage module. The reliability calculation process is as follows: subtract the difference of the components of the delay time precision dimension in the posterior mean vector from the delay time precision threshold, divide by the square root of the diagonal element of the delay time precision dimension in the posterior variance diagonal matrix to obtain the standardized distance value. Substituting the deviation value into the standard normal cumulative distribution function, the unreliability probability is calculated. Subtracting this unreliability probability from 1 yields the reliability. If the engineering distribution request includes constraint thresholds for assembly deviation or energy storage capacitor capacity, the reliability is calculated for each dimension separately, and the minimum value is taken as the final reliability. All detonators in the allocable state are arranged from highest to lowest reliability, and the required number of detonators at the top of the list are selected to form a candidate allocation list. It is then checked whether the lowest reliability value in the candidate allocation list is greater than or equal to 1 minus the maximum acceptable risk level. If this condition is met, the allocation method is confirmed. The plan is written into the distribution block. If the conditions are not met, the current inventory cannot meet the demand under the specified risk level. Each detonator in the candidate allocation list is divided into four risk level labels: green (excellent) for reliability greater than or equal to 99.5%, blue (good) for reliability between 97% and 99.5%, yellow (concern level) for reliability between 95% and 97%, and red (warning level) for reliability below 95%. After the allocation is confirmed, the project number, distribution timestamp, reliability value and risk label are written into the distribution block of the corresponding detonator in the life cycle data chain storage module.

[0014] Preferably, after the distributed detonators have completed the detonation operation during the usage phase, the detonation control system records the measured value of the actual delay time error for each detonator. The measured value of the actual delay time error is equal to the absolute value of the actual measured delay time minus the designed delay time. The lifecycle data link storage module receives the measured value of the actual delay time error through the edge data adapter and writes it into the usage feedback block of the corresponding detonator. Simultaneously, it records the successful or failed detonation status and the operation completion timestamp. The Bayesian state estimation engine reads the measured value of the actual delay time error from the usage feedback block and compares it with the component of the delay time accuracy dimension in the posterior mean vector recorded in the distribution block for the detonator. The prediction deviation value is calculated, which is equal to the actual measured delay time error minus the component of the delay time precision dimension in the posterior mean vector at the time of distribution. Model parameter calibration can be triggered through two independent paths: the first path is to trigger model parameter calibration when the absolute value of the prediction deviation value exceeds a preset calibration trigger threshold of 0.8 μs. The historical calibration dataset consists of paired samples composed of the actual measured delay time error values ​​recorded in the feedback block of all distributed detonators and the components of the delay time precision dimension in the posterior mean vector at the corresponding distribution time. Each sample in the calibration dataset contains a pair of predicted and measured values. When the number of paired samples in the historical calibration dataset is less than a preset minimum... When the sample size is 20, the model parameter calibration of the first path is not triggered. The loss function is constructed with the goal of minimizing the mean square error of the prediction on the historical calibration dataset. The expression of the target constructed loss function is the summation over all historical calibration samples, where each term equals the actual measured value of the delay time error z minus the square of the difference between the predicted mean μ and the predicted θ based on the decay model parameter θ. The decay model parameters are updated using the gradient descent algorithm with the goal of minimizing the target constructed loss function. The learning rate is set to 0.01, and the maximum number of iterations is set to 100. The second path triggers the rapid calibration of the decay model parameters when the number of samples in the divergence buffer queue reaches a preset cumulative number of 10. There are two triggering paths. Operating independently without blocking each other, after calibration, the updated attenuation model parameters are written back to the Bayesian state estimation engine through the edge data adapter, replacing the currently effective attenuation model. All subsequent state transition calculations of detonators in the stock are performed based on the updated attenuation model parameters. The risk quantification distribution and matching module recalculates the reliability of all detonators in the stock based on the updated probability distribution and updates the risk labels. If the reliability of a batch of detonators drops from above 97% to below 97% due to model correction, a batch-level early warning notification is automatically generated and written into the usage feedback block to form a complete traceability record. Thus, the closed-loop calibration process from feedback at the usage stage to reverse update of the state of all detonators in the stock is completed.

[0015] Beneficial effects This invention provides a method and system for full-cycle data management of electronic detonators. It has the following beneficial effects: This invention transforms the quality status of each detonator from a deterministic numerical value into a normal probability distribution characterized by mean and variance. This enables the quantitative differentiation of individual quality differences among detonators within the same batch. The reliability calculated based on the probability distribution replaces the traditional threshold judgment, allowing the distribution decision to consider both the quality estimate and the estimation uncertainty. This solves the problems of treating detonators with different quality statuses within the same batch equally, stockpiling near-expiry detonators, and misusing low-quality detonators in high-precision projects in the existing unified shelf-life management system. It provides an individualized risk quantification basis for warehousing zoning and outbound distribution.

[0016] This invention quantifies the consistency between observation results and prior predictions of the attenuation model using an information purity index. When discrepancies exist, the information propagation intensity is automatically reduced and stored in a discrepancy buffer queue, preventing the spread of false confidence among detonators in the same batch due to model parameter deviations. The model parameter calibration triggered by the discrepancy buffer queue and the reverse reassessment of all detonators in stock ensure that the calibration results promptly influence subsequent management decisions. By using information value scores, limited sampling resources are guided to the detonator position with the greatest cognitive uncertainty and closest to the distribution boundary, achieving a balance between inspection efficiency and distribution reliability. Attached Figure Description

[0017] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is an overall flowchart of the electronic detonator full-cycle data management method and system described in this invention. Detailed Implementation

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention provides a method and system for full-cycle data management of electronic detonators. The method includes a civil explosives company producing 500 electronic detonators with batch number B2025001. After production, each electronic detonator undergoes coding and quality inspection via a coding and testing terminal. The coding and testing terminal acquires a multi-dimensional quality feature vector for each electronic detonator, including measurements of four dimensions: assembly deviation, curing degree, energy storage capacitor capacity, and clock oscillator frequency. Taking electronic detonator B2025001-001 as an example, the assembly deviation measurement obtained by the coding and testing terminal is 0.12 mm, and the curing degree... The measured values ​​were 95.3%, the energy storage capacitor capacity was 47.2 μF, and the clock oscillator frequency was 32.0 kHz. Simultaneously, based on the accuracy levels of the measuring equipment for each dimension, the standard deviations for assembly deviation were determined to be 0.005 mm, curing degree was 0.02%, energy storage capacitor capacity was 0.5 μF, and clock oscillator frequency was 0.1 kHz. An initial normal probability distribution was constructed based on these four measurements, with the mean vector being a 4-dimensional vector containing components of 0.12, 95.3, 47.2, and 32.0, and the diagonal elements of the variance diagonal matrix being 0.005. 2 0.02 2 0.5 2 0.1 2 All off-diagonal elements are set to zero; the mean vector and the variance diagonal matrix are written into the production block of the detonator corresponding to number B2025001-001 in the life cycle data chain storage module, and the unique identifier, production batch number B2025001 and production completion timestamp are recorded at the same time; the above process is repeated for all 500 electronic detonators in batch B2025001 to complete the construction of all initial normal probability distributions.

[0020] After the 500 electronic detonators of batch B2025001 were received into the warehouse, they were allocated to three storage locations: C3, C4, and C5. Storage location C3 held 200 detonators, C4 held 200 detonators, and C5 held 100 detonators. The Bayesian state estimation engine used a preset decay model as its computational basis, starting with the initial normal probability distribution recorded in each detonator production block. It performed a state transition operation once for each set of temperature and humidity sampling data received. The preset initial value of the activation energy parameter Ea in the preset decay model was 1.0 eV, and the reference temperature T0 was 293 K. Based on batch number B2025... Taking the electronic detonator 001-001 as an example, when the temperature and humidity sensor collects that the temperature of the C3 compartment is 25℃ and the relative humidity is 60%, the Bayesian state estimation engine calculates the temperature acceleration factor and humidity diffusion coefficient, and multiplies the two to determine the mass decay increment vector for the current time step. The mean vector of the previous time step is subtracted from the mass decay increment vector to obtain the transition mean vector. The variance diagonal matrix of the previous time step is added to the preset process noise covariance matrix to obtain the transition variance diagonal matrix, which is written into the environmental block of the detonator with the number B2025001-001 in the life cycle data chain storage module.

[0021] After several days of warehousing, the quality management instruction specifies that electronic detonators numbered B2025001-017 in batch B2025001 be subject to quality sampling inspection. The measured delay time of the detonator is 2.3 μs, obtained using a delay accuracy measurement device, and the measured energy storage capacitor capacity is 46.8 μF, obtained using a capacitance measurement device. These two measured values ​​form an observation vector, and the diagonal elements of the measurement error covariance matrix are 0.5 squared and 0.3 squared, respectively. The Bayesian state estimation engine retrieves data from the lifecycle data chain storage module. The environmental block corresponding to detonator B2025001-017 reads the transition mean vector and transition variance diagonal matrix of the current time step as prior distribution parameters. After constructing the observation matrix, the Kalman gain matrix is ​​calculated using the Bayesian update formula, and then the posterior mean vector and posterior variance diagonal matrix are obtained. The diagonal elements of the posterior variance diagonal matrix are all reduced compared to the prior variance diagonal matrix. The update result is written into the sampling block of detonator B2025001-017 in the life cycle data chain storage module, and the detonator is marked as the sampled detonator.

[0022] The batch association update module then performs a reverse confidence propagation update based on information purity; and calculates the information purity index k of the sampled detonators. i , will k i Batch correlation coefficient ρ ijNonlinear modulation is performed to obtain the confidence propagation temperature coefficient. Since detonators numbered B2025001-017 and B2025001-018 to B2025001-040 are both located in warehouse C3 and belong to the same batch B2025001, after calculation using four layers of rules, the confidence propagation temperature coefficient of this batch of detonators is greater than the preset update threshold of 0.1, thus forming an affected detonator set. Mean shift and variance compression are performed on each detonator in the affected detonator set, and the updated mean vector and variance diagonal matrix are written into their respective sampling blocks.

[0023] After completing the above sampling update, the information value-driven inspection scheduling module recalculates the information value score for all detonators in stock. Taking electronic detonator B2025001-088 as an example, the sum of the diagonal elements of its current variance diagonal matrix is ​​0.38. Dividing this by the preset baseline variance sum of 0.260425 yields a standardized uncertainty index of approximately 1.46. Since the accuracy requirement threshold recorded in the most recent distribution request for this detonator is 2μs, the component of the delay time accuracy dimension in its posterior mean vector is 2.4μs. The absolute value of 0.4μs is divided by 10% of the accuracy requirement threshold, i.e., 0.2μs, to obtain a threshold sensitivity index of 2.0. The standardized uncertainty index of 1.46 is multiplied by the threshold sensitivity index of 2.0 and then by the preset comprehensive scaling factor of 100 to obtain an information value score of approximately 292. After normalization, it falls into the priority sampling queue. All detonators with information value scores greater than the preset queuing threshold of 10 are arranged in descending order of scores to form a priority sampling queue with a maximum length of 32 detonators. The sampling operation is performed sequentially according to the queue order and is dynamically updated.

[0024] When a high-precision demolition blasting project is received, requiring 30 detonators with a delay time accuracy threshold of 2μs and an acceptable maximum risk level of 1%, the risk quantification and distribution matching module iterates through all available detonators in the inventory and calculates the reliability of each detonator. Taking electronic detonator B2025001-017 as an example, the delay time accuracy threshold of 2μs is subtracted from the delay time accuracy dimension component of the posterior mean vector (2.1μs), resulting in a difference of 0.1μs. This difference is then divided by the posterior variance diagonal matrix. The standardized distance value is obtained by taking the square root of the diagonal element of the delay time precision dimension. This value is then substituted into the standard normal cumulative distribution function to calculate the unreliability probability, which is then subtracted from 1 to obtain the reliability. All detonators in the allocable state are sorted from highest to lowest reliability, and the top 30 are selected to form a candidate allocation list. The lowest reliability value in the list is checked to see if it is greater than or equal to 99%. If it is, the allocation scheme is confirmed. After classifying each detonator in the list into four risk levels, the project number, distribution timestamp, reliability value, and risk label are written into their respective distribution blocks.

[0025] After the 30 distributed electronic detonators have completed their detonation during the usage phase, the detonation control system records the measured value of the actual delay time error for each detonator. Taking electronic detonator B2025001-017 as an example, the measured value of the actual delay time error is 0.9 μs. The lifecycle data link storage module receives the measured value through the edge data adapter and writes it into the usage feedback block. The Bayesian state estimation engine compares the measured value of 0.9 μs with the component of the delay time accuracy dimension of the posterior mean vector recorded at the time of distribution, which is 2.1 μs. The absolute value of the prediction deviation is calculated to be 1.2 μs, exceeding the preset calibration trigger threshold. At 0.8μs, the model parameter calibration process for the first path is triggered. With the goal of minimizing the prediction mean square error on the historical calibration dataset, the activation energy parameter and diffusion coefficient parameter in the decay model are updated using the gradient descent algorithm. The updated parameters are written back to replace the currently effective decay model through the edge data adapter. All subsequent state transition operations of the detonators are performed based on the updated decay model parameters. The risk quantification distribution and matching module recalculates the reliability of all detonators based on the updated probability distribution and writes the timestamp of the calibration event, the prediction deviation value, and the updated decay model parameters into the feedback block to form a complete traceability record.

[0026] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for full-cycle data management of electronic detonators, characterized in that, include: S1 creates a block-based lifecycle data chain for each electronic detonator that is append-only and not modified, including production blocks, warehousing blocks, environment blocks, sampling blocks, distribution blocks, and usage feedback blocks. Each block records the corresponding stage data and uncertainty measures. S2, construct an initial normal probability distribution based on the multi-dimensional quality feature vector and measurement standard deviation collected by the coding detection terminal, and represent it with the initial mean vector and the initial variance diagonal matrix, and write it into the production block along with the production batch number and the production completion timestamp; S3, based on the temperature and humidity time series data, perform state transition operation with the currently effective decay model, subtract the mass decay increment vector from the mean vector of the normal probability distribution to obtain the transition mean vector, add the variance diagonal matrix to the process noise covariance matrix to obtain the transition variance diagonal matrix, and write it into the environmental block; S4, Receive the observation vector of quality sampling inspection, update the observation vector with the transition mean vector and transition variance diagonal matrix in the environment block based on Bayesian update, obtain the posterior mean vector and posterior variance diagonal matrix, and write them into the sampling inspection block; S5, receive an engineering distribution request that includes a precision requirement threshold and an acceptable maximum risk level, and calculate the reliability of the stocked electronic detonators in an allocable state based on the posterior mean vector and the posterior variance diagonal matrix. The reliability is calculated by subtracting the difference of the delay time precision dimension components in the posterior mean vector from the precision requirement threshold, dividing by the square root of the variance of the delay time precision dimension in the posterior variance diagonal matrix, and then substituting it into the standard normal cumulative distribution function. S6. Arrange the electronic detonators in the allocable state according to their reliability, select the required number of electronic detonators from the top engineering distribution requests to form a candidate allocation list, check whether the minimum reliability value meets the acceptable maximum risk level constraint, if it does, confirm the allocation scheme and write it into the distribution block.

2. The method for full-cycle data management of electronic detonators according to claim 1, characterized in that: The specific process of performing state transition calculation based on the currently effective attenuation model described in step S3 is as follows: Temperature and humidity time-series data are collected through environmental sensors. The temperature data in the temperature and humidity time-series data is substituted into the Arrhenius acceleration factor function to calculate the temperature acceleration factor. The humidity data is substituted into the diffusion coefficient function to calculate the humidity diffusion coefficient. The product of the temperature acceleration factor and the humidity diffusion coefficient is determined as the mass attenuation increment vector for the current time step. The mass attenuation increment vector is a multi-dimensional vector, with each dimension corresponding to the mass degradation increment in each dimension of the multi-dimensional mass feature vector described in step S2. The process noise covariance matrix described in step S3 is a diagonal matrix, with the diagonal elements reflecting the increase in uncertainty in each dimension at each time step due to unobserved random degradation factors. If a Bayesian update occurs between two state transition operations, the next state transition operation will start with the updated posterior mean vector and posterior variance diagonal matrix, rather than the previous transition result. After the state transition operation is completed, the temperature and humidity time series data, temperature acceleration factor, humidity diffusion coefficient, transition mean vector, and transition variance diagonal matrix will be written into the environmental block.

3. The method for full-cycle data management of electronic detonators according to claim 2, characterized in that: After obtaining the posterior mean vector and posterior variance diagonal matrix in step S4, a batch association update step is also included: determining the batch association coefficient between the electronic detonators that have undergone quality sampling in step S4 and the electronic detonators in the same batch in the warehouse. The same batch refers to the production batch number written into the production block in step S2. The batch association coefficient is determined according to the following rules: a basic association coefficient is set between electronic detonators with the same production batch number, and zero is set between different production batch numbers; on this basis, warehouse association correction is performed according to whether they are stored in the same or adjacent warehouse number, and transportation association correction is performed according to whether they have experienced the same transportation batch number and whether the vibration record exceeds the standard. Then, it is multiplied by the attenuation factor that decreases over time after the completion of joint production. The batch association coefficients after each layer of correction are truncated to between the upper and lower limits. The difference between the diagonal elements of the transition variance diagonal matrix and the posterior variance diagonal matrix in step S4 is taken as the information quantity. The information quantity is multiplied by the batch correlation coefficient and distributed proportionally according to the number of affected electronic detonators in the same batch. The mean vector of each electronic detonator in the same batch is shifted towards the update direction of the electronic detonators that have undergone quality sampling in step S4. The corresponding information allocation amount is subtracted from the variance diagonal matrix of each electronic detonator in the stock and written into the sampling block of its corresponding lifecycle data chain.

4. The method for full-cycle data management of electronic detonators according to claim 3, characterized in that: In the batch association update step, the information purity index of the electronic detonator that has undergone quality sampling inspection in step S4 is first calculated. The determined batch association coefficient is nonlinearly modulated with the information purity index to obtain the confidence propagation temperature coefficient. The modulation rule is that the larger the information purity index, the smaller the confidence propagation temperature coefficient. When the information purity index is zero, the confidence propagation temperature coefficient is equal to the batch association coefficient. When the information purity index approaches 1, the confidence propagation temperature coefficient approaches 0. In the batch association update step, the confidence propagation temperature coefficient is used instead of the batch association coefficient to perform information allocation and mean offset; When the information purity index exceeds the preset divergence threshold, the observation is marked as a suspicious observation, and the transfer mean vector and transfer variance diagonal matrix, observation vector, posterior mean vector and posterior variance diagonal matrix of the electronic detonator that have undergone quality sampling in step S4 are stored in the divergence buffer queue.

5. The method for full-cycle data management of electronic detonators according to claim 4, characterized in that: When the number of samples in the divergence buffer queue reaches the preset cumulative number, the fast calibration process of the currently effective decay model in step S3 is triggered: using the information purity index of each sample in the divergence buffer queue as the weight, a weighted loss function is constructed, and the activation energy parameter and diffusion coefficient parameter in the currently effective decay model are iteratively updated through the gradient descent algorithm. After calibration, the divergence buffer queue is cleared. After calibration, a reverse reassessment is triggered. The updated attenuation model is used to re-execute the state transition operation in step S3 from the time of entry into the warehouse. For each electronic detonator in the warehouse, the gradient of the normal probability distribution with respect to the parameters in the currently effective attenuation model is calculated. Only electronic detonators with gradient absolute values ​​exceeding a preset sensitivity threshold are fully recalculated. Electronic detonators with gradient absolute values ​​below the threshold are approximated using a first-order Taylor expansion. After the recalculation is completed, the current posterior mean vector and posterior variance diagonal matrix of each electronic detonator are updated. Each sample includes prior distribution parameters, observation vectors, and posterior distribution parameters.

6. The method for full-cycle data management of electronic detonators according to claim 5, characterized in that: After step S4 and before step S5, there is a step of generating a priority sampling queue: calculating the information value score for each electronic detonator in stock, the information value score being determined by a combination of uncertainty reduction potential index and decision influence weight index; if the electronic detonator has never participated in the engineering distribution request in step S5, the decision influence weight index is calculated using a preset benchmark accuracy threshold as the default accuracy requirement threshold; electronic detonators with information value scores greater than the preset queuing threshold are arranged in descending order of score to form a priority sampling queue; after each sampling is completed according to step S4, the information value score of the affected electronic detonators is immediately recalculated, and the priority sampling queue is dynamically updated.

7. The method for full-cycle data management of electronic detonators according to claim 6, characterized in that: In step S6, each electronic detonator in the candidate allocation list is divided into four risk level labels according to its reliability value, and each risk level label corresponds to a preset reliability value range. When the engineering distribution request in step S5 includes constraint thresholds for assembly deviation and energy storage capacitor capacity, the reliability of each dimension is calculated based on the components of the corresponding dimension in the posterior mean vector and posterior variance diagonal matrix in step S4, and the minimum value among the reliability of each dimension is taken as the final reliability. If the lowest reliability value in the candidate allocation list in step S6 does not meet the acceptable maximum risk level constraint, then feedback is sent to the engineering team that the current inventory cannot meet the demand under the specified risk level; after allocation confirmation, the project number, distribution timestamp, reliability value and risk level label are written into the distribution block of the lifecycle data chain.

8. The method for full-cycle data management of electronic detonators according to claim 7, characterized in that: The parameter calibration of the currently effective attenuation model in step S3 is also triggered through the following path: After the distributed electronic detonators have completed the detonation operation during the use phase, the actual measured value of the delay time error is obtained and written into the use feedback block; the measured value is compared with the component of the delay time accuracy dimension in the posterior mean vector in step S4, and the prediction deviation value is calculated. When the absolute value of the prediction deviation value exceeds the preset calibration trigger threshold, the parameter calibration of the currently effective attenuation model in step S3 is triggered, so that the gradient descent iterative update is performed with the goal of minimizing the prediction mean square error of the updated attenuation model on the historical distribution data; After calibration, the updated attenuation model replaces the currently effective attenuation model described in step S3, and all subsequent state transition calculations for all electronic detonators in the inventory are performed based on the updated attenuation model. At the same time, the reliability of all electronic detonators in the inventory is recalculated based on the updated normal probability distribution. If the reliability of a batch of electronic detonators drops from the qualified state in the candidate allocation list in step S6 to the unqualified state due to model correction, a batch-level early warning notification is automatically generated and written into the usage feedback block.

9. A full-cycle data management system for electronic detonators, used to implement the full-cycle data management method for electronic detonators as described in claim 1, characterized in that, include: The lifecycle data chain storage module is used to create a unique lifecycle data chain for each electronic detonator from the time of production coding. The lifecycle data chain adopts a block-based structure that only adds and does not modify, and includes six types: production block, warehousing block, environment block, sampling block, distribution block and usage feedback block. Each block records the data and uncertainty measurement of the corresponding stage. The Bayesian state estimation engine is used to perform state transition operations based on the currently active decay model to obtain the transition mean vector and the transition variance diagonal matrix, and to receive the observation vector and perform Bayesian updates to obtain the posterior mean vector and the posterior variance diagonal matrix. The batch association update module determines the batch association coefficient between the sampled electronic detonator and the electronic detonators in the same batch in the warehouse, and performs reverse confidence propagation based on the information purity index to update the probability distribution of the electronic detonators in the same batch in the warehouse. The information value-driven inspection and scheduling module is used to calculate the information value score for each electronic detonator in stock, and to generate and dynamically update the priority sampling queue. The risk quantification distribution and matching module is used to calculate the reliability based on the posterior mean vector and the posterior variance diagonal matrix after receiving the engineering distribution request, and to perform risk quantification distribution and matching. An edge data adapter is used to convert the native data format into a standardized data block format and write it into the lifecycle data chain storage module, and to send the measured feedback data during the usage phase back to the Bayesian state estimation engine to achieve model parameter calibration.

Images