A neural network model-based hole-by-hole initiation method

Abstract

The application provides a detonation method based on a neural network model hole-by-hole detonation technology, and the application optimizes the delay time and selects the special detonator, eliminates the micro-difference time coincidence, realizes the true sense of hole-by-hole detonation, and improves the crushing effect by more than 30%. A plurality of types of reliability models are established, the minimum branch minimum value criterion is used to quickly locate the weak link of the network, and the network misfire rate is reduced to below 0.5%. The application is based on a 12-dimensional input parameter BP neural network model, the prediction error of the blasting vibration parameter is controlled within 5%, and the safety of surrounding buildings is effectively ensured. The neural network prediction is deeply integrated with the detonation network design, construction implementation, the detonation parameter dynamic optimization is realized, various rock properties and blasting area conditions are adapted, and the universality is strong. The standardized network structure and the fast connection block are used, the network laying time is reduced, large-scale area blasting is adapted, and the human operation failure rate is reduced.

Description

Technical Field

[0001] This invention relates to the field of blasting technology, and more specifically, to a detonation method based on a hole-by-hole detonation technique using a neural network model. Background Technology

[0002] Sequential detonation technology, a core technology in modern engineering blasting, utilizes a combination of in-hole and out-of-hole millisecond delay detonators to achieve sequential detonation of blast holes. This optimizes blasting fragmentation effects, reduces blasting vibration hazards, and enhances blasting safety. Traditional sequential detonation network design relies heavily on manual experience, lacking quantitative basis for delay time selection and network topology design. This can easily lead to problems such as simultaneous detonation of blast holes, overlapping micro-delay times, and mis-detonation, severely impacting blasting quality and construction safety.

[0003] Existing non-electric millisecond detonators have inherent deviations in their delay times, and the detonation delay of detonating cords is affected by wiring length. Furthermore, traditional initiation network reliability analysis only provides qualitative assessments of equipment and structure, failing to accurately predict blasting vibration parameters. This results in significant challenges in controlling blasting vibrations, making it difficult to meet the blasting safety requirements of areas near buildings and precision instruments. Simultaneously, traditional initiation networks exhibit poor delay time matching; the nominal delay time difference between detonators in some sections is an integer multiple of the delay time of external relay detonators, easily leading to simultaneous detonation of multiple boreholes, violating the original design principle of sequential detonation.

[0004] Neural network models possess powerful nonlinear fitting and multi-factor prediction capabilities, enabling them to accurately fit the complex mapping relationships between blasting vibration parameters and various influencing factors, thus overcoming the shortcomings of low accuracy and poor adaptability in traditional empirical formulas. Currently, there is no initiation method that deeply integrates BP neural network models with hole-by-hole initiation network design, reliability analysis, and vibration parameter prediction, making it impossible to achieve intelligent and precise control of the entire hole-by-hole initiation process. Therefore, this paper proposes an initiation method based on a neural network model for hole-by-hole initiation technology. Summary of the Invention

[0005] The purpose of this invention is to address the problems identified in the existing background technology. To achieve the above-mentioned objective, this invention provides the following technical solution: a detonation method based on a neural network model for sequential detonation, comprising the following steps: Step 1: Conduct an analysis of the detonation network delay time, establish a formula for calculating the actual delay time of the borehole, and select suitable in-hole and out-of-hole delay detonator segments to avoid the risk of synchronous detonation of boreholes. Step 2: Based on the number of free surfaces and the hole layout of the blast zone, design a hole-by-hole initiation network with a corresponding topology and limit the included angle of the detonation direction to ensure the reliability of the detonation. Step 3: Construct a reliability model for the detonation network, and combine it with the reliability index of the detonation element to quantitatively evaluate the relative reliability of the detonation network using specified criteria; Step 4: Build a three-layer BP neural network vibration parameter prediction model, determine the number of nodes and corresponding parameters in the input layer, hidden layer, and output layer, and train the model using on-site blasting data to achieve accurate vibration parameter prediction. Step 5: Based on the BP neural network prediction results and reliability assessment conclusions, dynamically adjust the detonation parameters to complete the detonation network laying and hole-by-hole detonation implementation; The formula for calculating the actual delay time of the blast hole is as follows: In the formula: Let n be the actual delay time for the nth borehole in the mth row. The nominal extension period for the detonator inside the hole. The nominal delay time for the external relay detonator is given, and x is the number of the relay detonator ignition sequence.

[0006] As a preferred technical solution of the present invention, in step one, the hole-by-hole detonation delay deployment mode includes two types: one is to install detonators of the same level in the hole, and to use low-level detonators to relay detonation between holes and between rows; the other is to use a combination of inter-hole relay, inter-row level difference or inter-row relay and inter-hole level difference delay mode.

[0007] As a preferred technical solution of the present invention, in step one, the core principle for avoiding synchronous detonation is: the difference between the nominal delay time of any two segments of the in-hole delay detonator is not an integer multiple of the nominal delay time of the external relay detonator; the external relay detonator uses 9ms, 17ms, 25ms, 42ms, 65ms, and 100ms series of surface delay detonators.

[0008] As a preferred technical solution of the present invention, in step two, the adaptation rules of the per-hole detonation network topology and the detonation zone conditions are as follows: a V-shaped detonation network is used for a single free-face detonation zone, an oblique detonation network is used for a double free-face detonation zone, and a triangular per-hole detonation network is used for a triangular per-hole detonation zone; and the angle between the row detonation direction and the column detonation direction is greater than or equal to 90°.

[0009] As a preferred technical solution of the present invention, in step three, the detonation network reliability model includes four basic models: series, parallel, parallel-series, and series-parallel.

[0010] As a preferred technical solution of the present invention, in step three, the reliability index of the detonating element is as follows: the reliability of the detonating cord detonator is 0.9612, the reliability of the detonating cord is 0.9683, and the reliability of the reflective four-way connector is 0.9843, with a confidence level of 0.95 for each; the detonating cord is measured in units of 20m, and the detonating cord is measured in units of 50m.

[0011] As a preferred technical solution of the present invention, in step four, the three-layer BP neural network model is constructed as follows: the input layer is set with 12 nodes, and the corresponding parameters are the most recent charge amount, the maximum charge amount, the total charge amount, the height difference of the most recent charge, the height difference of the maximum charge, the average height difference, the detonation time of the most recent charge, the detonation time of the maximum charge, the shortest micro-difference time between holes, the longest micro-difference time between holes, the distance of the most recent charge, and the distance of the maximum charge; the output layer is set with 3 nodes, and the corresponding parameters are the vibration velocity amplitude, the vibration dominant frequency, and the vibration duration; the number of hidden layer nodes is set to 25.

[0012] As a preferred technical solution of the present invention, in step four, the BP neural network model needs to re-collect the blasting data of the corresponding medium for iterative training in the region where the characteristics of the blasting medium change, and then carry out vibration parameter prediction.

[0013] As a preferred technical solution of the present invention, in step one, the selection rules for the optimal micro-difference delay time are as follows: 15-30ms for hard rock, 20-46ms for medium-hard rock, and 50-70ms for soft rock; the optimal delay time per unit hole spacing is 3-8ms / m, and the optimal delay time per unit row spacing is 8-15ms / m.

[0014] As a preferred technical solution of the present invention, in step five, for branches with weak reliability of the detonation network, the reliability of the branches is improved by using parallel double detonators and redundant double detonating cords; for irregular blasting areas, 9ms, 17ms, and 25ms surface delay detonators are used for network fine-tuning, and a special connecting block is used to realize the rapid connection of the surface network.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Completely avoid the risk of synchronous detonation: By optimizing the delay time and selecting special detonators, the overlap of micro-delay time is eliminated, realizing true hole-by-hole detonation, and improving the crushing effect by more than 30%.

[0016] 2. Precisely quantify detonation reliability: Establish multiple types of reliability models and use the minimum branch minimum value criterion to quickly locate weak links in the network, reducing the network misfire rate to below 0.5%.

[0017] 3. Significantly improved vibration forecast accuracy: Based on a BP neural network model with 12-dimensional input parameters, the prediction error of blasting vibration parameters is controlled within 5%, effectively ensuring the safety of surrounding buildings.

[0018] 4. Intelligent control throughout the entire process: The neural network prediction is deeply integrated with the design of the detonation network and the construction implementation to achieve dynamic optimization of detonation parameters, adapt to various rock types and blasting zone conditions, and has strong versatility.

[0019] 5. Significantly improved construction efficiency: The adoption of a standardized network structure and quick-connect blocks reduces network laying time, adapts to large-scale blasting, and reduces the rate of human error. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the inter-hole force provided by the present invention; Figure 2 This is a schematic diagram of the inter-row force provided by the present invention; Figure 3 A schematic diagram illustrating the relationship between inter-hole delay time and explosive block size provided by the present invention; Figure 4 This is a schematic diagram of single-free-plane V-shaped sequential detonation provided by the present invention; Figure 5 A schematic diagram of the double free-plane oblique line hole-by-hole initiation network provided by the present invention; Figure 6 This is a schematic diagram of the per-hole network for triangular hole arrangement provided by the present invention; Figure 7 The reliability logic diagram of the series system provided by this invention; Figure 8 The reliability parallel system logic diagram provided for this invention; Figure 9 The reliability and serial connection system logic diagram provided by this invention; Figure 10 The reliability serial-parallel system logic diagram provided by the present invention; Figure 11 This is a schematic diagram of the mine inclined micro-delay initiation network system provided by the present invention; Figure 12 This is a schematic diagram of the BP network structure provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely illustrates some embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention. It should be noted that, in the absence of conflict, the embodiments and features and technical solutions in the embodiments of the present invention can be combined with each other. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] Example 1: A detonation method based on a neural network model for hole-by-hole detonation, including hole-by-hole detonation network design and reliability analysis. The hole-by-hole detonation network design includes two scenarios: in-hole delay and out-of-hole delay. Depending on the blasting scale, and with the aid of blasting equipment, these two delay methods can be flexibly used to design various complex detonation networks. When designing the detonation network, it is necessary to understand and analyze the nominal delay time of non-electric millisecond detonators and the reasonable micro-delay time of blasting, select appropriate surface delay detonators, and perform reliability calculations on the designed detonation network.

[0024] Analysis of the delay time of the detonation network: To achieve the detonation technology of hole by hole, from the perspective of network design, there are usually two ways to achieve it: one is to install detonators of the same level in all holes, and use low-level detonators to relay between holes and between rows; the other is to relay between holes and between rows, or relay between rows and between holes.

[0025] Inter-hole force-fed and inter-row step difference refers to placing detonators of different segments in holes with different numbers of rows, while placing detonators of the same segment in holes within the same row. The holes within the same row are separated by external force-fed detonators, while the rows are separated by the difference in the nominal delay time of each segment. Inter-row force-fed and inter-hole step difference refers to placing detonators of different segments sequentially in holes within each row. The rows are separated by external force-fed detonators, while the holes are separated by the difference in the nominal delay time of each segment.

[0026] The time required for a given blast hole to go from the detonation of each ignition detonator to the detonation of the detonator inside the hole can be calculated using the following formula: T mn =t mn +x·t d (4-1); where: T mn t represents the actual delay time (time required for the detonator in the nth borehole of the first row to detonate from the moment the detonator is ignited.) mn The nominal delay time for the detonator of the m-th borehole in the m-th row; t d denoted as the nominal delay time for the external detonator; x is the sequential number of the relay detonator ignited in the m-th row and n-th borehole.

[0027] Assume the actual delay time of a certain blast hole in the network is T. mn1 The actual delay time for the other blast hole is T. mn2 Then, according to formula (4-1), we can obtain: T mn1 =t mn1 +x1·t d (4-2); T mn2 =t mn2 +x2·t d (4-3); Subtracting the two equations, we get: T mn2 -T mn1 = (+(x2-x1)·t) d (4-4); As can be seen from (4-4), when t mn2 -t mn1 With (x2-x1)·t d When the absolute values ​​are equal but the signs are opposite, T mn2 -T mn1 =0, that is, T mn2 =T mn1 In other words, if the actual delay time of the two blast holes is equal, then the two blast holes will detonate simultaneously.

[0028] It is obvious that (x2-x1)·t d It is t d an integer multiple of t, if t mn2 -t mn1 With (x2-x1)·t d If the absolute values ​​of t are equal, then t mn2 -t mn1 The absolute value of t is also t d This means that the difference in the nominal delay time between two borehole detonators must be an integer multiple of the nominal delay time of the relay detonator outside the borehole. In other words, if the difference in the nominal delay time between any two detonators used for borehole delay in a network is an integer multiple of the nominal delay time of the relay detonator, then two or more boreholes in this network may detonate simultaneously. This contradicts the original design intent of the network and fails to meet the requirements of blasting safety.

[0029] The non-electric millisecond detonators commonly used in my country are generally divided into 20 segments, with the 1-15 segment detonators being the most commonly used. Their nominal delay times are as follows: segment 1 (0ms), segment 2 (25ms), segment 3 (50ms), segment 4 (75ms), segment 5 (110ms), segment 6 (150ms), segment 7 (200ms), segment 8 (250ms), segment 9 (310ms), segment 10 (380ms), segment 11 (460ms), segment 12 (550ms), segment 13 (650ms), segment 14 (760ms), and segment 15 (880ms). Analyzing the time differences above, it's easy to see that the nominal delay time differences between segments 12 and 6, 13 and 6, 8 and 7, 12 and 7, 13 and 7, 12 and 8, 11 and 9, 14 and 9, 15 and 10, 14 and 11, and 13 and 12 are all multiples of 25ms (2 segments), 50ms (3 segments), or 75ms (4 segments). However, the nominal delay time differences between segments 9 and 7, 15 and 12, and 14 and 13 are multiples of 110ms (5 segments). If detonators with 6 or more segments are used for in-hole delay and detonators with 5 or fewer segments are used for external relay, some holes will always detonate simultaneously, compromising network safety.

[0030] The above only lists the nominal delay time. Due to manufacturing processes, the nominal delay time of detonators may vary. Furthermore, considering the time required for the detonating cord to propagate the detonation (the detonation wave propagates stably in the pipe at a low detonation velocity of 1600–2000 m / s), taking 2 m / ms, it becomes clear that during use, as the length of the detonating cord increases, the delay time cannot be effectively controlled, and the probability of micro-delay time overlaps will greatly increase.

[0031] When using the sequential detonation technology, since the delay time series of its surface delay detonators are 9ms, 17ms, 25ms, 42ms, 65ms, and 100ms, there will be no overlap of micro-delay times, thus enabling true sequential detonation.

[0032] Hole-by-hole detonation network: To reduce blasting vibration at critical locations, Tatsuya Heshibo et al. [i]A method for optimizing the delay time considering the initiation time difference of electric detonators is proposed. By simulating the superposition of vibrations generated by explosions at various boreholes at a certain point using a calculation program, the peak particle velocity (PPV) at that point can be calculated. Using the relationship between the PPV at that point and the delay time, the micro-delay time for obtaining the minimum PPV can be easily determined. However, the actual delay time varies due to errors in the initiation time. For this reason, the expected composite PPV depends not only on the delay time but also on the sensitivity of the PPV. Therefore, the selection of the optimal interval time for micro-delay blasting should be based on the optimal time interval and should include several alternative suboptimal time intervals. In addition, the error in the initiation time should be fully considered. Tatsuya Heshibo et al. also proposed a new design concept: composite micro-delay blasting. Composite micro-delay blasting uses electronic detonators for initiation, which can further reduce the PPV at a specified point. The vibration time history and PPV estimated based on this concept show that the blasting vibration reduction effect is quite ideal.

[0033] For open-pit bench blasting, both electric and non-electric initiation remain the most common methods. Non-electric initiation systems include plastic detonating cord initiation systems, detonating cord relay initiation systems, and gas detonating cord initiation systems. The plastic detonating cord initiation system is the most widely used. Its composition is similar to that in my country, for example, a relay initiation network where high-level detonators are used inside the borehole and low-level detonators are used outside. The boreholes in the same section can be single-hole or multi-hole. Initiation methods include oblique, V, and U-shaped patterns. Meanwhile, many projects still use sequential initiation, where subsequent rows of boreholes are detonated in the designed sequence before the first row is fully detonated, creating a visually striking oblique initiation pattern.

[0034] The large-scale blasting blasting at the Wyoming open-pit coal mine in the United States utilized the aforementioned plastic detonating cord relay detonation network. To facilitate open-pit mining, the mine blasted and stripped the overburden layer, using a blasting method to eject 30% of the blasting debris and remove the remaining 70%. The overburden layer was 20–60 m thick, and the coal seam was 20 m thick. The blasting parameters were: borehole diameter F 200–300 mm; hole depth 20–60 m; resistance line 8–14 m; hole spacing 10–13 m; number of rows 6–8; approximately 100 boreholes per row; inter-hole delay time 25–30 ms; inter-row delay time 100–200 ms; explosive charge per blast ranging from 1400 to 3000 tons; and a blast volume of 700,000 m³. 3 ~1.5 million m 3 .

[0035] In the Hong Kong Special Administrative Region, open-pit blasting controls the hazards of blast vibrations by strictly controlling the maximum amount of explosives used in each delay phase of the blast. The Hong Kong Mining Department, based on years of accumulated blast vibration monitoring data, has developed a formula for predicting blast vibrations: PPV=644(D / Q1 / 2 ) -122 (4-5); In the formula: PPV is the maximum vibration velocity of the mass point, mm / s; D is the distance from the seismograph point to the blasting zone, m; Q is the maximum charge per delay segment, kg.

[0036] During blasting operations, the Hong Kong Mining Department, based on the different structures surrounding the site, listed the maximum allowable charge for each delay interval at various distances, which was strictly adhered to during blasting operations. Simultaneously, non-electric detonating cord micro-delay initiation systems were widely adopted, primarily the Swedish Unidit system (500ms in-hole delay, with inter-hole delay achieved through different connection methods of ground-based detonators). The main advantage of the Unidit system is that it allows for an unlimited number of boreholes to be detonated in a single operation, and the charge for each delay interval (minimum 8ms) can be very small, thus increasing the scale of blasting while reducing blast vibration. Because it allows for sequential detonation of each borehole, it effectively improves blasting quality and controls blast hazards. Furthermore, since the Unidit system requires fewer delay intervals to produce, it significantly reduces the workload in the factory.

[0037] Currently, foreign products widely adopt multi-layered high-strength plastic detonating cords, with different colors used for different applications, making them easy to identify and reducing the chance of errors. Furthermore, foreign countries place great emphasis on the design and improvement of plastic connecting blocks for surface detonating detonators, resulting in very few instances of detonation interruption due to loose connecting blocks.

[0038] According to dynamic load theory and the requirement of minimum blasting vibration, extensive research and practice have proven that for open-pit deep-hole blasting, when the minimum resistance line is 6-10m, the following minerals and rocks should be selected for micro-differential interval time as follows: (1) Granite, olivine, gabbro, diorite, quartzite, etc., T Z =15~30ms; (2) Serpentine, hard limestone, porphyry, sandstone, etc., T Z =20~46ms; (3) Tough and relatively soft minerals and rocks, such as magnesite, gypsum, marl, etc., T Z =50~70ms. According to data provided by Hanukaev: for medium-hard rock types, the time required for the first detonation charge to form a new free face after the second detonation charge is generally 25~60ms.

[0039] In recent years, based on the above theories and a large number of experiments, foreign research institutions have proposed the concepts of unit hole spacing delay time and unit row spacing delay time. They believe that when the optimal delay time between holes in the same row is selected within the range of 3 to 8 ms / m and the optimal delay time between rows is selected within the range of 8 to 15 ms / m, the best crushing effect can be achieved.

[0040] To maximize the interaction between two blast holes and achieve the best fragmentation effect, the optimal delay time between blast holes in the same row is 3-8 milliseconds per meter of hole spacing. In actual blasting operations, the specific value used depends on the properties of the rock at the blasting site.

[0041] During the field industrialization test, based on the rock properties of different blasting zones and with full consideration of the safety of the blasting network, the surface delay hole inter-delay time was designed to be 25ms and 42ms, and the row inter-delay time was designed to be 42ms, 65ms and 100ms. The final application was based on the actual field conditions (including hole network parameters, rock properties, etc.).

[0042] Definition and classification of initiation network reliability: The ability of an initiation network to complete its predetermined functions, such as initiation and delay, under specified conditions and within a specified time is called reliability. Since completing the predetermined function is a random event, it can be quantitatively expressed using probability, which is called reliability degree.

[0043] The reliability of an initiation network can be categorized into relative reliability and absolute reliability depending on the specified conditions. Relative reliability refers to the reliability of the initiation network under normal design, installation, and use conditions, considering only certain individual factors such as the reliability of the initiating elements and the design form of the network system. It is an index value used to compare different network types. Absolute reliability, on the other hand, considers not only the aforementioned conditions but also human error in network design, installation, and use; it is the network system reliability that takes into account the combined effects of various influencing factors and represents the actual reliability of the initiation network.

[0044] The time specified in the definition refers to the design reference period from the start of network installation to the completion of its intended function. Generally speaking, for the same object, the longer the specified time, the lower the reliability. However, for detonation networks, since the entire time from the start of installation to the completion of its intended function is very short, the changes in the components of the network within this short period are minimal and have little impact on network reliability. Furthermore, since detonation networks are one-time operating systems, they differ in reliability from some systems that require continuous operation. Therefore, the time factor can be ignored in the reliability calculation of detonation networks. For underwater and humid areas where the installation time of the detonation network is long, the calculation should be limited to the specific condition that the detonating elements do not fail within the effective time.

[0045] Factors affecting the reliability of non-electric initiation network systems: The reliability of non-electric initiation network systems is determined by the following four factors: (1) Reliability of the initiating element itself: The initiating element includes detonating cord detonator, detonating cord, detonating cord and its connecting elements. Since the initiating element is the material basis of the initiation network, the reliability of the initiating element itself undoubtedly has a direct impact on the reliability of the initiation network system.

[0046] (2) Detonation network layout: The relative reliability varies depending on the network design. The relative reliability of the network is determined by the designer based on factors such as the quality and type of the detonating equipment. A reasonable design approach can achieve a high relative reliability for the entire detonation network. Conversely, an inappropriate design may result in a very low relative reliability, or even partial misfires.

[0047] (3) Network Laying Technology: The network laying technology is an important factor affecting the reliability of non-electric detonating detonators. It has a significant correlation with the reliability of the network. To reduce the impact of network laying on its reliability, a reliable construction method must be established, including: reliable methods and measures to ensure the safe detonation of non-electric detonating cords by detonating detonators; technological methods and technical measures to prevent detonators from damaging other non-electric detonating cords in the network when they are detonated; and technical means to ensure that different sections of adjacent branches do not produce overlapping or serial sections.

[0048] (4) Construction technology management: Advanced detonation technology and methods need to be implemented by people. Through a large number of field investigations, it has been found that there are many cases of misfires caused by human error in network operation, resulting in missing or incorrect connections. Therefore, the influence of this factor cannot be ignored.

[0049] (1) Determination of the interface of the detonation network system: The detonation network refers to the system formed by the connection of the various components that constitute the detonation network, which does not include the detonating charge. The interface of the system in the borehole extends to the detonator that detonates the detonating charge.

[0050] (2) Determination of the attributes and scope of the factors affecting the detonation network: Of the four influencing factors mentioned above, only the network laying type is deterministic; the other three are random factors. Furthermore, since both network laying technology and construction management are random events resulting from human behavior, there are currently no scientific quantitative methods available for calculating network reliability, and therefore these will not be discussed in this paper. Therefore, the reliability calculation described in this paper is actually the calculation of the reliability of the detonation network determined by the two factors of the detonating element and the network laying type. Clearly, this is the relative reliability of the detonation network, and based on this, the reliability of different network types is compared, and their respective reliability evaluations are made.

[0051] The network laying method and the reliability of the network system components themselves are two independent events. The former is a deterministic event, while the latter is a random event. Based on the above calculation of the reliability of the detonation network, it can be reduced to calculating the network reliability determined by each component in different network laying methods.

[0052] 1.1 Establishment of the reliability model of the detonation network system: When performing reliability analysis and reliability calculation on the detonation network system, it is necessary to establish a reliability model of the system. The so-called reliability model of the system refers to the reliability diagram of the system and one or more mathematical expressions, which can represent the relationship between the reliability of the system and the reliability of its detonating elements. The establishment of the system reliability model can be carried out in the following steps: (1) Divide the system into several elements according to the concepts of elements and system; (2) Draw the reliability logic diagram of the system according to the failure modes of the elements; (3) Establish the reliability mathematical model of the system based on the reliability logic diagram.

[0053] A system with cascaded reliability is a system in which the failure of any one subsystem will cause the failure of the entire system, and the reliability of each subsystem is independent of the others.

[0054] The reliability logic diagram of a series system borrows the drawing method from the series circuit diagram, but the reliability logic diagram of the network differs from the structure diagram. The former represents the logical relationship between the system state and the normal state of the components, while the latter represents the physical relationship between the system and the components.

[0055] In a reliable cascaded system: ; In the formula: R s —Reliability of series systems; R i —The reliability of the i-th subsystem; Reliable parallel system: Definition: A system is called a parallel reliability system if it fails only when all its subsystems fail, and the reliability of each subsystem is independent of the others. Logic diagram of a parallel reliability system.

[0056] In a reliable parallel system, ; In the formula: R s —Reliability of parallel systems; R i —The reliability of the i-th subsystem; Reliable parallel-serial systems: In reliable parallel-serial systems ; In the formula: R s —Reliability of parallel-series systems; R ij —Reliability of the ij-th subsystem: Reliable series-parallel systems.

[0057] In reliable series-parallel systems, ; In the formula: R s —Reliability of series and parallel systems; R ij —The reliability of the ij-th subsystem; The four reliability models mentioned above are the most basic reliability models commonly used in detonation network systems.

[0058] The reliability of an initiating element refers to its ability to perform its intended function under specified operating conditions within its specified storage life. The probabilistic quantitative characterization of reliability is called reliability degree.

[0059] Because the term reliability has statistical properties—inferring the probability of future events based on past observations; using sample statistical parameters to predict parameters of a population with the same distribution—it falls under the category of statistical inference. Therefore, a reliability estimate can only be valid under a certain level of confidence.

[0060] Methods for estimating the reliability of detonating elements: The reliability index of detonating elements is estimated using the counting method. When a certain performance of the detonating element is tested using the counting method, only whether it is qualified or unqualified is judged. The standard for qualification is determined according to the technical specifications. The obtained data are the total number of tests n (i.e., sample size) and the number of qualified items s (or the number of failures f). The lower limit of the qualified rate in the population estimated at a confidence level r is the reliability R.

[0061] Since the reliability test of the detonating element is a success-failure system test, when the total batch size N is large enough (N≥10n), the number of successes S (or the number of qualified products) among the n tested detonating elements is a random variable that conforms to a binomial distribution, and the reliability R is determined by equation (2-5) or (2-6).

[0062]

[0063] If n, f, and r are known, R can be solved using equations (2-5) or (2-6), but this is very difficult to calculate in practice. Therefore, an iterative method can be used on a computer to solve the problem.

[0064] Reliability indicators for several non-electric initiation elements and detonation transmission nodes: Element and junction names Reliability indicator Confidence Detonating-cable detonator 0.9612 0.95 Detonating cable 0.9683 0.95 Detonating cord 0.999 0.90 Reflecting cross 0.9843 0.95 Detonating-cord-to-detonating-cable transmission point 0.9834 0.90 Detonating-cable-to-detonating-cable transmission point 0.9943 0.95 Note: The reliability values ​​for detonating cords in the table are based on a unit length of 20m; the reliability values ​​for detonating cords are based on a unit length of 50m.

[0065] Criterion for evaluating the relative reliability of non-electric initiation network systems: using the minimum value of the smallest branch.

[11] As a criterion for evaluating the relative reliability of non-electric initiation network systems.

[0066] The criterion for the minimum value of the minimum branch is: in a certain type of detonation network system, the minimum reliability of the last minimum branch (or borehole) in all its subsystems is taken as the relative reliability of the detonation network system.

[0067] The following examples will demonstrate the basis and rationale for adopting this criterion.

[0068] Example: An open-pit mine, considering vibration reduction, adopts a borehole-to-bore differential inclined initiation method, using a non-electric detonating cord initiation network system. The number of boreholes in the j-th row is m. j All detonators are dual-shot parallel detonating cord millisecond detonators, and each hole uses a highly reliable detonating cord for detonation. Assume the reliability of the detonator is R, and the reliability of the detonating cord is R'. D Determine the reliability R of the detonation network system. S .

[0069] Each subsystem of its detonation network is a parallel-series network. According to its calculation principle, the reliability R of the j-th subsystem is... Sj for:

[0070] Where: (i=1, 2, ..., m) j (k=1, 2, ..., n). In the formula: R sjG R sjz — These represent the relative reliability of the design detonation transmission for the main and branch lines at row j, respectively; R K R i — These represent the relative reliability of each detonation point on the main detonation line and the branch detonation line at row j.

[0071] The relative reliability R of the detonation network system S : ; Its symbolic meaning is the same as before; As can be seen from equation (2-7), when the number of detonators in the network is large, (Because 0) <R<1,0<R D <1), which means that in large-scale micro-differential blasting using tens of thousands of detonators, it is almost impossible to ensure that every borehole explodes without any misfires. Therefore, studying this approximately zero R... SThat would be meaningless. When conducting reliability evaluations of non-electric initiation network systems, an appropriate structural hierarchy should be selected. Specifically, the key factors affecting network reliability should be identified, which is determined by the unique characteristics of initiation network system research.

[0072] For non-electric initiation network systems, the key element is the minimum value of the last smallest branch in all subsystems of the gray network system. This is because the smallest branch is the basic unit constituting the initiation network, and the minimum relative reliability of all subsystems calculated based on it is precisely the weakest link in the entire initiation network system. Its reliability characteristics largely determine the overall reliability characteristics of the network, and its reliability level is reflected numerically and accurately.

[0073] Furthermore, the reliability calculations for the detonation network in this paper are all relative reliability calculations. The resulting reliability is not the true reliability of the detonation network. However, for the reliability evaluation of the network system, it is still acceptable to select the network form with high reliability by comparing relative reliability. Therefore, when designing a detonation network, it is only necessary to calculate the relative reliability according to this criterion, and it is not necessary to calculate the reliability of the entire network.

[0074] Of course, the method of using this criterion to evaluate the reliability of detonation networks is not yet perfect. Because many factors influence the reliability of detonation networks, how to more scientifically and organically integrate this criterion with the entire network system is a question worthy of future exploration.

[0075] Example 2: Neural networks are highly nonlinear dynamic systems. Although the structure and function of each neuron are very simple, the behavior of a network system composed of a large number of neurons is extremely rich and complex. [ii] It has an adaptive and learning process, which uses data from network training samples to find quantitative relationships between system inputs and outputs, thereby completing system prediction.

[0076] A backpropagation (BP) network is a feedforward network composed of nonlinear transformation units and has backpropagation capabilities. It is one of the most commonly used neural network models. It not only has input layer nodes and output layer nodes, but also hidden layer nodes, which can be one or more layers.

[0077] The BP (Backpropagation) network algorithm is a self-learning algorithm based on gradient descent. Input information is processed from the input layer through hidden layer units and then transmitted to the output layer. The state of each neuron only affects the state of the next layer's neurons. If the desired output cannot be obtained at the output layer, backpropagation is initiated, returning the error signal along the original connection path. By modifying the weights of each neuron, the error signal is minimized.

[0078] Suppose the network has P input samples and the input information vector is X. k(k = 1, 2, ..., P) Expected output information vector T k (k=1,2,…P). The algorithm steps of a BP network can then be described as follows: (1) Initialize the network by randomly assigning initial values ​​to the connection weights and thresholds.

[0079] (2) Input the learning samples and calculate the input and output of each layer according to equations (1) and (2).

[0080] ; In the formula: W ij For nodes i With nodes j Connection rights between, subscript i For the parent node, the subscript j To and i The adjacent next-level node; For nodes j The threshold; The function applied to the node can be... S It can be a nonlinear function or a linear function.

[0081] (3) Calculate the total network error according to formula (3). E , such as total error E ≥ e Or individual error Proceed to step 4; otherwise, the requirements are met, and the network output is executed.

[0082] ; (4) Calculate the gradient correction error of each layer using equation (4). Then the weights are adjusted using equation (5).

[0083] ; In the formula: To learn step size; For gradient correction error, ; c These are the iteration coefficients.

[0084] Establishment of the BP network forecasting model; Determining the input layer nodes; Engineering blasting is a complex multi-factor control system. Many parameters affect vibration, and the vibration parameters vary depending on the blasting method and initiation method.

[0085] Based on mechanical theory and engineering blasting practice experience, and building upon previous empirical formulas, the following functional expression can be used to express the relationship between blasting vibration parameters and influencing factors: ; In the formula: These are blasting vibration parameters, composed of multiple indicators; There are many factors that affect vibration. i =1~ n ; The vibration expression function is an uncertain function.

[0086] For such a complex system with multiple factors influencing its output and multiple indicators, conventional mathematical methods and experimental techniques are insufficient to determine its functional relationships. The successful application of neural network theory in complex systems controlled by multiple factors provides an effective approach to solving such problems.

[0087] By using vibration influencing factors as input layer nodes and blasting vibration parameters as output layer nodes, and determining reasonable intermediate layer nodes (hidden layer nodes) and connection weights between the input and output layers of the network, a BP network model for blasting vibration prediction can be established.

[0088] For sequential detonation, due to the abundance of micro-delay stages, the influencing factors of blasting vibration are more complex than those of traditional micro-delay blasting. These factors can be considered from three aspects: medium factors, blast source factors, and measuring point factors.

[0089] Medium factors directly affect the propagation and attenuation of seismic waves, such as medium structure and properties. Obtaining medium characteristic parameters from the field is complex. When using empirical formulas to predict seismic parameters, medium factors must be considered. However, neural network models can eliminate the influence of the medium by training the model to adapt to changes in the characteristics of the blasting medium. In other words, in areas where the blasting medium characteristics are similar, a neural network model can accurately obtain target information after a certain amount of training. Once the medium characteristics change, the model needs to be retrained for the corresponding medium before it can make predictions. Therefore, when establishing a neural network model, the blasting medium factor can be ignored. However, in blasting areas where the medium characteristics change significantly, the model needs to be trained before it can make predictions.

[0090] The factors influencing vibration parameters include the explosive charge, charge location, and initiation time. In sequential initiation, the explosive charge in each hole is not necessarily the same, and calculating the explosive charge for all holes is tedious if the blasting scale is large. Based on engineering experience, we can consider some typical explosive charges that have a major impact on vibration parameters to build a model. Through the analysis of the mechanism of sequential initiation, this model mainly considers three explosive charge parameters: the charge of the nearest point to the measuring point, the maximum charge, and the total charge. The charge location can be reflected by the height difference between the measuring point and the charge. Based on theory and practical experience, we need to consider three location parameters: the height difference of the nearest charge, the height difference of the maximum charge, and the average height difference. In addition to considering the initiation time of the nearest and maximum charges, the time difference between holes also needs to be considered. The use of high-precision millisecond detonators allows for greater freedom in designing the inter-hole differential time for sequential detonation. The inter-hole differential time is uncertain in different scenarios and can have multiple inter-hole differential times. For convenience, the inter-hole differential time is considered here in terms of two factors: the shortest inter-hole differential time and the longest inter-hole differential time.

[0091] The main factors to consider when determining the measuring point are the relative distances from the measuring point to the blasting zone. Combining the theory of shock wave propagation and practical experience, the two parameters to be considered here are the distance to the nearest explosive charge and the distance to the maximum explosive charge.

[0092] In summary, when establishing a BP network forecasting model, the vibration influencing factors that need to be considered are: the nearest charge amount, the maximum charge amount, the total charge amount, the nearest charge height difference, the maximum charge height difference, the average height difference, the nearest charge detonation time, the maximum charge detonation time, the shortest inter-hole differential time, the longest inter-hole differential time, the nearest charge distance, and the maximum charge distance, totaling 12 influencing factors. These factors determine 12 nodes for the model's input layer.

[0093] Determining the output layer nodes; Regardless of the type of blasting, previous "Blasting Safety Regulations" mostly used vibration velocity as the blasting safety evaluation index. In recent years, many scholars have conducted in-depth research on the safety evaluation of blasting vibrations, and almost unanimously agree that using vibration velocity alone to measure blasting safety is incomplete. Research results indicate that using vibration velocity amplitude, dominant vibration frequency, and vibration duration as indicators for a comprehensive evaluation of blasting vibration safety is more reasonable and accurate than using a single evaluation index. Therefore, when establishing the BP network prediction model, the research focuses on these three vibration parameters; that is, the model's output layer has three nodes: vibration velocity amplitude, dominant vibration frequency, and vibration duration.

[0094] Determining hidden layer nodes; Rebert proved that, within a certain range, a BP network with one hidden layer can complete the mapping from any n-dimensional to m-dimensional system. Therefore, when establishing a per-hole initiation vibration parameter prediction model, a BP network with one hidden layer is selected.

[0095] There is no theoretical formula for calculating the number of hidden layer nodes; it can only be selected based on empirical formulas.

[46] The following two are commonly used empirical formulas for calculating hidden layer nodes: ; In the formula: n The number of hidden layer nodes; m This refers to the number of factors influencing blasting vibration, i.e., the number of nodes in the input layer. l This represents the number of vibration parameters that need to be predicted, i.e., the number of nodes in the output layer. a It is a constant between 1 and 10.

[0096] Data shows that the more hidden layer nodes a network has, the better its prediction performance. However, too many hidden layer nodes significantly increase the computational load of the model, severely impacting the network's convergence speed. Therefore, it is necessary to minimize the number of hidden layer nodes while maintaining prediction performance. Some researchers have indicated that the number of hidden layer nodes should generally not be less than the number of nodes in the input and output layers. After comparing and analyzing empirical formulas, the final recommended number of hidden layer nodes is 25.

[0097] At this point, the structure of the model has been determined. It is a 3-layer BP network model containing 12 input nodes, representing the most recent charge amount, the maximum charge amount, the total charge amount, the height difference between the most recent charge and the maximum charge, the average height difference, the initiation time of the most recent charge and the initiation time of the maximum charge, the shortest time difference between holes, the longest time difference between holes, the distance between the most recent charge and the distance between the maximum charge; 3 output nodes, representing the vibration velocity amplitude, the dominant vibration frequency, and the vibration duration; and 25 hidden layer nodes used to ensure the accuracy of network calculations.

[0098] Clearly, the connection weights between the input layer and the hidden layer are a 12×25 matrix, and the connection weights between the hidden layer and the output layer are a 25×3 matrix.

[0099] The mechanism of blasting vibration is extremely complex and difficult to describe using theoretical models. However, the highly nonlinear characteristics of neural networks can effectively solve this problem, fully demonstrating the superiority of neural network theory. Calculation and analysis results show that the BP network model is more accurate than empirical formulas in predicting blasting vibration parameters under various geological conditions, considering more influencing factors and predicting more comprehensive parameters—something traditional mathematical models and empirical formulas cannot achieve. After training, the BP network model is very convenient and flexible to apply; by changing some of the model's influencing factors, vibration parameter predictions can be achieved under various conditions.

[0100] Before application, BP network models need to be trained. The more samples used in training, the more accurate the forecast results will be. This requires collecting a certain amount of raw data and field monitoring results based on the vibration influencing factors considered by the model, in order to ensure the accuracy of the forecast model.

[0101] The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described herein. Although the present invention has been described in detail with reference to the above embodiments, the present invention is not limited to the specific embodiments described above. Therefore, any modifications or equivalent substitutions to the present invention, as well as all technical solutions and improvements that do not depart from the spirit and scope of the invention, are covered within the scope of the claims of the present invention.

Claims

1. A detonation method based on a neural network model for hole-by-hole detonation, characterized in that, Includes the following steps: Step 1: Conduct an analysis of the detonation network delay time, establish a formula for calculating the actual delay time of the borehole, and select suitable in-hole and out-of-hole delay detonator segments to avoid the risk of synchronous detonation of boreholes. Step 2: Based on the number of free surfaces and the hole layout of the blast zone, design a hole-by-hole initiation network with a corresponding topology and limit the included angle of the detonation direction to ensure the reliability of the detonation. Step 3: Construct a reliability model for the detonation network, and combine it with the reliability index of the detonation element to quantitatively evaluate the relative reliability of the detonation network using specified criteria; Step 4: Build a three-layer BP neural network vibration parameter prediction model, determine the number of nodes and corresponding parameters in the input layer, hidden layer, and output layer, and train the model using on-site blasting data to achieve accurate vibration parameter prediction. Step 5: Based on the BP neural network prediction results and reliability assessment conclusions, dynamically adjust the detonation parameters to complete the detonation network laying and hole-by-hole detonation implementation; The formula for calculating the actual delay time of the blast hole is as follows: In the formula: Let n be the actual delay time for the nth borehole in the mth row. The nominal extension period for the detonator inside the hole. The nominal delay time for the external relay detonator is given, and x is the number of the relay detonator ignition sequence.

2. The detonation method based on a neural network model for hole-by-hole detonation according to claim 1, characterized in that, In step one, the sequential detonation delay deployment mode includes two types: one is to install detonators of the same level in the hole, and use low-level detonators to relay detonation between holes and between rows; the other is to use a combination of inter-hole relay, inter-row level difference, or inter-row relay and inter-hole level difference delay mode.

3. The detonation method based on a neural network model for hole-by-hole detonation according to claim 2, characterized in that, In step one, the core principle for avoiding synchronous detonation is that the difference between the nominal delay time of any two segments of the in-hole delay detonator is not an integer multiple of the nominal delay time of the external relay detonator; the external relay detonator uses 9ms, 17ms, 25ms, 42ms, 65ms, and 100ms series of surface delay detonators.

4. The detonation method based on a neural network model for hole-by-hole detonation according to claim 1, characterized in that, In step two, the rules for adapting the topology of the hole-by-hole detonation network to the blasting zone conditions are as follows: a V-shaped detonation network is used for single-free-face blasting zones, an oblique-line detonation network is used for double-free-face blasting zones, and a triangular hole-by-hole detonation network is used for triangular hole-by-hole blasting zones; and the angle between the row detonation direction and the column detonation direction is greater than or equal to 90°.

5. The detonation method based on a neural network model for hole-by-hole detonation according to claim 1, characterized in that, In step three, the detonation network reliability model includes four basic models: series, parallel, parallel-series, and series-parallel.

6. The detonation method based on a neural network model for hole-by-hole detonation according to claim 5, characterized in that, In step three, the reliability indicators for the detonating elements are as follows: detonating cord detonator reliability 0.9612, detonating cord reliability 0.9683, and reflective four-way connector reliability 0.9843, all with a confidence level of 0.95; the detonating cord is measured in 20m units, and the detonating cord is measured in 50m units.

7. The detonation method based on a neural network model for hole-by-hole detonation according to claim 1, characterized in that, In step four, the three-layer BP neural network model is constructed as follows: the input layer has 12 nodes, with corresponding parameters being the most recent charge amount, the maximum charge amount, the total charge amount, the height difference between the most recent charge and the maximum charge, the average height difference, the detonation time of the most recent charge and the detonation time of the maximum charge, the shortest time difference between holes, the longest time difference between holes, the distance between the most recent charge and the distance between the maximum charge; the output layer has 3 nodes, with corresponding parameters being the vibration velocity amplitude, the dominant vibration frequency, and the vibration duration. The number of hidden layer nodes is set to 25.

8. The detonation method based on a neural network model for hole-by-hole detonation according to claim 7, characterized in that, In step four, the BP neural network model needs to re-collect blasting data of the corresponding medium for iterative training in areas where the characteristics of the blasting medium change, and then carry out vibration parameter prediction.

9. The detonation method based on a neural network model for hole-by-hole detonation according to claim 1, characterized in that, In step one, the optimal delay time selection rules are as follows: 15-30ms for hard rock, 20-46ms for medium-hard rock, and 50-70ms for soft rock; the optimal delay time per unit hole spacing is 3-8ms / m, and the optimal delay time per unit row spacing is 8-15ms / m.

10. The detonation method based on a neural network model for hole-by-hole detonation according to claim 1, characterized in that, In step five, for branches with weak reliability in the detonation network, the reliability of the branches is improved by using parallel double detonators and redundant double detonating cords; for irregular blasting areas, 9ms, 17ms, and 25ms surface delay detonators are used for network fine-tuning, and a special connecting block is used to realize rapid connection of the surface network.

Images