Non-disassembly verification method for industrial tool management device

Abstract

The application relates to a non-disassembly verification method for an industrial cutter management device, which comprises the following steps: obtaining a taking request containing a target cutter model identifier and calling a verification reference set; collecting a frequency offset curve generated by a packaging box crossing a multi-axis coil and a shielding timestamp sequence of an optical-electric node, combining a crossing speed to convert a transient magnetic impedance modulus value curve after vector synthesis into a spatial domain electromagnetic integral quantity; obtaining an axial passing length according to the shielding timestamp sequence; collecting a total mass of the packaging box; substituting the spatial domain electromagnetic integral quantity into a regression model to deduce a predicted metal mass, and then combining the total mass to calculate an estimated packaging mass; and comprehensively judging based on the spatial domain electromagnetic integral quantity, the axial passing length, the total mass and the estimated packaging mass and outputting a verification signal. The application has the advantages that the internal cutter model and integrity can be accurately verified without damaging the packaging, and environmental interference factors such as packaging damp and falling posture are eliminated.

Description

Technical Field

[0001] This application relates to the field of consumables management, and in particular to a non-unpacking verification method for industrial tool management equipment. Background Technology

[0002] In machining production environments, carbide end mills, as high-value consumable tools, range in price from hundreds to thousands of yuan per piece, making them a core control target for workshop material management. Traditional tool management relies on a manual tool room issuance model: operators collect tools from the tool room with work orders, and the administrator visually verifies the model and quantity before manually issuing them. This model faces problems in large workshops, such as high record-keeping error rates, inability to retrieve tools outside of working hours, difficulty in tracing tool issuance, and the inability to digitally track tool consumption.

[0003] Therefore, automated industrial tool dispensing systems have been introduced into tool management scenarios. These systems achieve end-to-end digital control of tools through a combination of compartmentalized storage, identification authentication, and usage records. However, carbide end mills are stored and distributed in individually sealed packages. Upon release, the system cannot directly determine the model and integrity of the contents of the package through visual or tactile means, posing a risk of mis-dispensing or substandard products.

[0004] Existing automated dispensing equipment primarily relies on two solutions for tool verification: barcode scanning or single-dimensional mass weighing. Barcode scanning only verifies the packaging label and cannot detect whether the metal parts inside the packaging match the label. Single-dimensional mass weighing faces two types of failure scenarios: cutting fluid mist in the workshop causes paper packaging to absorb moisture over a long period, and the packaging weight fluctuates with humidity, leading to misjudgments based on fixed mass thresholds; when different models of tools have similar masses, mass information alone cannot distinguish between them. Summary of the Invention

[0005] In order to simultaneously perceive the material properties and geometric dimensions of the metal parts inside the packaging without opening the packaging, and to cross-verify them with quality information, this application provides a non-open packaging verification method for industrial tool management equipment.

[0006] This application provides a non-unpacking verification method for industrial tool management equipment, which adopts the following technical solution: A non-unpacking verification method for industrial tool management equipment includes: S1. Obtain a requisition request, the requisition request containing the model identifier of the target tool; retrieve the verification benchmark set corresponding to the model identifier from the benchmark database according to the model identifier, the verification benchmark set containing a regression model and a judgment threshold; S2. Collect the multi-channel frequency offset curves generated by the multi-axis coils in the detection area of ​​the packaging box, as well as the occlusion timestamp sequence of multiple photoelectric nodes; determine the crossing speed of the packaging box in the detection area based on the occlusion timestamp sequence; perform vector synthesis on the multi-channel frequency offset curves to obtain the transient magnetoresistance modulus curve; combine the crossing speed to convert the transient magnetoresistance modulus curve into a spatial domain electromagnetic integral quantity; S3. Based on the occlusion timestamp sequence, obtain the axial length of the packaging box; S4. Collect the total mass of the packaging box at the weighing position; wherein the weighing position is set in a preset drop endpoint area; S5. In response to successfully obtaining the total mass, substitute the spatial domain electromagnetic integral into the regression model in the verification benchmark set to obtain the predicted metal mass; based on the difference between the total mass and the predicted metal mass, obtain the estimated packaging mass; S6. Based on the spatial domain electromagnetic integral, the axial passage length, the total mass, and the estimated packaging mass, and by comparing with the judgment threshold in the verification benchmark set, output a verification pass signal or an abnormal signal.

[0007] Optionally, the multi-axis coil includes a first axis coil, a second axis coil, and a third axis coil orthogonally fixed to the outer wall of the non-metallic guide channel, and a first LC oscillation circuit, a second LC oscillation circuit, and a third LC oscillation circuit electrically connected to the first axis coil, the second axis coil, and the third axis coil, respectively; the non-metallic guide channel is an insulated hollow tubular channel that penetrates the detection area, the detection area is the inner cavity space of the non-metallic guide channel, and the packaging box falls through the inner cavity of the non-metallic guide channel and passes through the detection area; The process of collecting the multi-channel frequency offset curves generated by the multi-axis coil during the process of the packaging box passing through the detection area, as well as the occlusion timestamp sequence of multiple photoelectric nodes, includes the following sub-steps: S21. When the non-metallic guide channel is in an unloaded state, collect the three reference frequencies output by the first LC oscillation circuit, the second LC oscillation circuit, and the third LC oscillation circuit respectively; S22. Collect the real-time output frequencies of the first LC oscillation circuit, the second LC oscillation circuit, and the third LC oscillation circuit during the process of the packaging box passing through the detection area, as well as the occlusion timestamp sequence of the multiple photoelectric nodes; S23. Subtract the corresponding reference frequency from the real-time output frequency of the first LC oscillation circuit, the second LC oscillation circuit and the third LC oscillation circuit respectively to obtain the multi-channel frequency offset curve; S24. Perform time-synchronized root mean square calculation on the frequency offset of each path in the multi-path frequency offset curve to obtain the transient magnetoresistance magnitude curve; determine the crossing speed according to the obstruction timestamp sequence; and convert the transient magnetoresistance magnitude curve into the spatial domain electromagnetic integral quantity in combination with the crossing speed.

[0008] Optionally, S21 further includes: The three reference frequencies are re-acquired at preset calibration time intervals, and the corresponding stored values ​​in the reference database are replaced with the re-acquired three reference frequencies. If the sampling variance of any reference frequency exceeds the preset reference variance threshold, stop accepting requests and output a calibration anomaly alarm.

[0009] Optionally, in step S2, determining the crossing speed of the packaging box through the detection area based on the occlusion timestamp sequence and converting the transient magnetoresistance modulus curve into a spatial domain electromagnetic integral includes the following sub-steps: Obtain the occlusion timestamp sequence of N photoelectric nodes, where N is an integer not less than 4; Based on the fixed spacing between adjacent photoelectric nodes and the corresponding occlusion time interval, the instantaneous velocity of each of the N-1 sub-intervals is obtained; The transient magnetoresistance modulus curve is divided into N-1 segments according to the occlusion timestamp sequence. The time domain integral of each segment is multiplied by the instantaneous velocity of the corresponding sub-interval and then summed to obtain the spatial domain electromagnetic integral. The fixed spacing between adjacent photoelectric nodes is no greater than one-third of the axial coverage length of the detection area.

[0010] Optionally, in S3, the axial through length is obtained through the following steps: Based on the instantaneous velocity of each of the N-1 sub-intervals and the corresponding occlusion time interval, the path component of each sub-interval is obtained; The axial passage length of the packaging box within the detection area is obtained by summing the path components of each of the N-1 sub-intervals.

[0011] Optionally, S2 further includes: Using twice the time required for the maximum helical groove spacing of the target tool corresponding to the model identifier to pass through the detection area as the sliding window length, the maximum sliding value envelope filter is applied to the transient magnetoresistance modulus curve to obtain the envelope curve; In response to the envelope curve having a single peak, it is determined that the internal metal parts of the packaging box are a single continuous body; In response to the presence of multiple independent peaks in the envelope curve with an interval length greater than the length of the sliding window, the abnormal signal is output; wherein fluctuations in the envelope curve with amplitudes lower than a preset proportion of the amplitude of the main peak of the envelope curve are not included in the count of independent peaks.

[0012] Optionally, step S4 includes the following sub-steps: S41. Collect the continuous output of the weighing sensor located at the weighing position, and calculate the rolling variance of the continuous output within a preset duration; S42. In response to the rolling variance being lower than a preset stability threshold, the average of the continuous outputs within the preset duration is used as the total mass; S43. In response to the fact that the rolling variance is still not lower than the preset stability threshold after a preset timeout period has elapsed since the packaging box reached the weighing position of the weighing sensor, output a timeout exception signal and terminate the current verification process.

[0013] Optionally, in step S6, the determination threshold includes the allowable range of the feature ratio, the allowable range of the packaging quality tolerance, and the allowable range of the axial passing length; S6 includes the following sub-steps: In response to the fact that the ratio of the total mass to the electromagnetic integral in the spatial domain falls within the characteristic ratio allowable range, the estimated packaging mass falls within the packaging mass tolerance range, and the axial passing length falls within the axial passing length allowable range, the verification pass signal is output. If any condition is not met, the abnormal signal is output.

[0014] Optionally, prior to step S1, the method further includes a step of establishing the benchmark database, which includes the following sub-steps: For no less than a preset number of genuine cutting tool samples corresponding to the model identifier, data pairs of spatial domain electromagnetic integral quantity and actual metal mass are collected under the condition of covering multiple representative rolling posture angles. Using the spatial domain electromagnetic integral as the input variable and the actual metal mass as the output variable, a multinomial regression is performed to obtain the regression model, which includes model coefficients and corresponding upper bounds of model residuals.

[0015] In summary, this application includes at least one of the following beneficial technical effects: 1. By substituting the spatial domain electromagnetic integral into the regression model to deduce the predicted metal mass, and then subtracting the total mass from the predicted metal mass to extract the estimated packaging mass, the verification process achieves data decoupling between the internal metal entity and the external packaging material. This overcomes the defects in quality judgment caused by the long-term moisture absorption of paper packaging due to cutting fluid mist in the machining workshop and the random fluctuation of packaging weight with humidity, thus improving the accuracy of automated material dispensing verification.

[0016] 2. By acquiring the multi-channel frequency offset curves of multi-axis coils for time-synchronized vector synthesis, and combining the partitioned instantaneous velocity extracted by multiple photoelectric nodes, the time domain signal is converted into the spatial domain electromagnetic integral quantity. This eliminates the measurement error caused by the random change of the rolling posture and the drift of the friction coefficient of the guide channel when the packaging box falls freely in the guide channel, so that the extracted electromagnetic features always stably correspond to the real physical volume and material of the tool.

[0017] 3. By integrating spatial domain electromagnetic integral, axial through length, total mass, and estimated packaging quality to construct a multi-dimensional physical feature cross-verification system, the equipment can perform transparent comparison of the internal carbide end mills without removing the original factory sealed packaging, preventing fraudulent activities such as wrong delivery, missing delivery, or using low-value materials to impersonate high-value solid tools.

[0018] 4. By applying a sliding maximum envelope filter to the transient magnetoresistance modulus curve and analyzing the peak shape, the system can directly use physical signals to identify the geometric continuity of the metal parts inside the packaging box, thereby accurately distinguishing a continuous solid end mill from a bulk broken blade of the same weight, further improving the control dimension of tool outbound integrity. Attached Figure Description

[0019] Figure 1 A flowchart illustrating a non-unpacking verification method for an industrial tool management device according to some embodiments of this application is shown.

[0020] Figure 2 This is a schematic diagram of the detection channel in one embodiment of the present invention.

[0021] Figure label: 10. Grid outlet; 20. Non-metallic guide channel; 21. Inner cavity of the guide channel; 30a, First photoelectric node; 30b, Second photoelectric node; 30c, Third photoelectric node; 30d, Fourth photoelectric node; 41. First axis coil; 42. Second axis coil; 43. Third axis coil; 50. Receiving tray; 60. Weighing sensor; 70. Equipment base. Detailed Implementation

[0022] The present application will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the application and are not intended to limit the scope of the application.

[0023] Before detailing the specific implementation methods of this application, some terms are explained below. The target cutting tool can be a high-value metal cutting tool such as a carbide end mill or drill bit used in industrial production. The packaging box can be a paper box, a plastic liner, or a combination of both to wrap the target cutting tool. The detection area is the inner cavity of the non-metallic guide channel 20, which is an insulated hollow tubular channel penetrating the detection area. The packaging box falls through the inner cavity of the non-metallic guide channel 20 and passes through the detection area. The benchmark database is a set of model index parameters pre-collected and established for each model of cutting tool before the issuance verification process is executed. It includes the regression model and judgment threshold corresponding to each model.

[0024] The industrial tool management equipment includes a grid storage array, a pushing mechanism, a non-metallic guide channel 20, a detection area, a landing area, a weighing module, and a control module. The grid storage array consists of multiple rows and columns of independent grids, each storing one target tool in its original sealed packaging. The grid inlet faces the operator, and the bottom of the grid connects to the top inlet of the guide channel. The pushing mechanism is located on the inner wall of each grid and uses a pusher plate structure driven by an electromagnetic pusher or stepper motor. Upon receiving an outbound command, it horizontally pushes the packaging box within the grid to the inlet of the guide channel, allowing the packaging box to enter the non-metallic guide channel.

[0025] Figure 2 A schematic diagram of the detection channel according to some embodiments of this application is shown. For example... Figure 2 As shown, the detection channel is mainly composed of a non-metallic guide channel 20, which runs vertically through the inside of the equipment. After the packaging box P enters the inner cavity 21 of the guide channel from the grid outlet 10, it accelerates downward along the axis under the action of gravity, passes through the photoelectric speed measurement zone and the multi-axis electromagnetic detection zone in sequence, and finally reaches the receiving tray 50 at the bottom.

[0026] It should be noted that during the process of packaging box P entering the non-metallic guide channel 20 from the grid outlet 10, due to the action of the pusher plate of the pushing mechanism on the tail end of packaging box P, the front end of packaging box P first crosses the edge of the grid outlet 10 and loses support. At this time, the tail end of packaging box P is still in contact with the bottom surface of the grid, and the front end begins to sink under the action of gravity, causing packaging box P to generate an initial tumbling torque around the edge of the grid outlet 10. After the tail end of packaging box P completely leaves the bottom surface of the grid, packaging box P enters the inner cavity 21 of the guide channel with a non-zero angular velocity and continues to tumble during the subsequent descent. The intensity of the tumbling depends on the geometric relationship between the grid outlet 10 and the inlet of the non-metallic guide channel 20, the speed of the pusher plate, and the aspect ratio and mass distribution of packaging box P itself. Because the cross-sectional dimensions of the non-metallic flow channel 20 match the maximum cross-section of the packaging box P, the inner wall of the flow channel provides a geometric constraint on the tumbling amplitude of the packaging box P, limiting the maximum tilt angle of the packaging box P within the inner cavity 21 of the flow channel. This ensures that the falling trajectory of the packaging box P is always constrained within the axial channel range of the flow channel. The impact of the initial tumbling motion on each test stage within the detection area will be explained in the corresponding sections below.

[0027] The non-metallic guide channel 20 is an insulated hollow tubular channel extending vertically, made of polytetrafluoroethylene or ABS engineering plastic. Its cross-sectional dimensions match the maximum cross-section of the packaging box, and its inner wall is treated with a low-friction coating to ensure the packaging box can fall stably and accelerate under gravity. The non-metallic guide channel 20 covers three functional areas from top to bottom: a photoelectric velocity measurement area, a multi-axis electromagnetic detection area, and a landing point receiving area. Figure 2 As shown, the photoelectric velocity measurement area is located in the upper section of the non-metallic guide channel 20, with first photoelectric nodes 30a, second photoelectric nodes 30b, third photoelectric nodes 30c, and fourth photoelectric nodes 30d evenly arranged along the axial direction. When the packaging box P falls through it, it sequentially blocks the beams of each node, and the control module records the corresponding blocking timestamps t1, t2, t3, and t4. The fixed spacing ΔL between adjacent photoelectric nodes is no greater than one-third of the axial coverage length of the detection area. The transmitting and receiving ends of each node are fixed to both sides of the non-metallic guide channel 20. When the packaging box passes through, it sequentially blocks the beams, and the control module records the blocking timestamp sequence of each node. The multi-axis electromagnetic detection area is located in the middle section of the non-metallic guide channel 20, as shown... Figure 2 As shown, the multi-axis coil assembly 40 includes a first-axis coil 41, a second-axis coil 42, and a third-axis coil 43, which are orthogonally fixed to the outer wall of the non-metallic guide groove 20, and wound along the X-axis, Y-axis, and Z-axis directions, respectively. When the packaging box P carrying the internal metal parts passes through the inner cavity 21 of the guide groove, the metal parts cause changes in the magnetic field in the three orthogonal directions. Figure 2As shown by the horizontal arrows, the three coils respectively sense the changes in magnetic impedance along the corresponding axes and output frequency signals through their respective LC oscillation circuits. The three coils also sense the changes in magnetic impedance of the metal parts inside the packaging box in the X, Y, and Z axes, and convert the sensed signals into collectable frequency signals through the first, second, and third LC oscillation circuits, respectively.

[0028] like Figure 2 As shown, a receiving tray 50 is located below the bottom of the non-metallic guide channel 20. The bottom of the receiving tray 50 is connected to the weighing sensor 60 via an elastic buffer structure, and the weighing sensor 60 is fixed to the equipment base 70. The elastic buffer structure is used to absorb the impact energy when the packaging box P arrives, enabling the weighing sensor 60 to obtain a stable total mass reading after the oscillation decays. After passing through the photoelectric velocity measurement zone and the multi-axis electromagnetic detection zone above, the packaging box accelerates under the action of gravity to reach the receiving tray 50 and collides, then generates vertical elastic oscillations on the tray surface. The oscillation amplitude gradually decays with energy dissipation until it stops. The weighing sensor 60 of the weighing module can be a cantilever beam strain gauge weighing sensor 60, with a range covering 0 to 500g and a resolution not higher than 0.1g; it can also be a piezoelectric weighing sensor, with a higher dynamic response frequency, suitable for scenarios with large gaps and high impact speeds of the packaging box; it can also be a spring magnetic induction weighing sensor, which senses mass changes by causing the magnetic core to shift due to spring deformation. There are no restrictions here. The receiving tray 50 is connected to the force-bearing end of the weighing sensor 60 via an elastic link. An elastic buffer structure is provided on the bottom of the tray to absorb the impact energy when the packaging box arrives. This elastic buffer structure can be a rubber pad or a spring-damped shock absorber; the specific form should be designed to absorb the impact without affecting the stable weighing reading. After the packaging box arrives at the tray, the control module continuously collects the output of the weighing sensor 60. After confirming that the oscillation has sufficiently decayed using the rolling variance window criterion, it extracts the total mass reading and sends the total mass along with the spatial domain electromagnetic integral obtained from the electromagnetic detection area to the verification algorithm.

[0029] The target cutting tools in this application primarily refer to solid carbide end mills, including flat-end mills, ball end mills, and nose end mills, with shank diameters ranging from 4mm to 20mm and total tool lengths ranging from 50mm to 150mm. The base material of the solid carbide end mill can be tungsten-cobalt (WC-Co) cemented carbide, tungsten-titanium-cobalt (WC-TiC-Co) cemented carbide, or coated cemented carbide with a TiN or TiAlN coating applied to the surface of a tungsten-cobalt base using a physical vapor deposition process. The density of these materials ranges from approximately 13 g / cm³. 3 Up to 15.5 g / cm 3 All of these are far higher than the 7.8 g / cm³ of ordinary steel. 3Density characteristics form the physical basis for determining the ratio of electromagnetic integral quantity to mass. The cutting edge has 2 to 6 helical grooves distributed along the axial direction, with helix angles ranging from 30° to 45°. These helical grooves cause the cross-sectional area of ​​the cutting edge to change periodically along the axial direction, which is the fundamental reason for the geometric texture fluctuations in the electromagnetic integral signal, requiring processing through envelope filtering. The target cutting tool's packaging box has a rectangular structure, with an outer shell made of folded cardboard and an inner lining of foamed polyethylene or molded pulp trays. The tool shank is fixed by inserting it into the central positioning hole. The outer wall of the packaging is printed with a barcode and model number. The packaging box is entirely made of non-metallic material and contains no metal components. After entering the electromagnetic detection area, it does not produce a measurable shift in the resonant frequency of the LC oscillation circuit. The frequency shift signal only comes from the metal body of the target cutting tool inside. This is the physical premise for the electromagnetic differential scheme to achieve non-unpacking detection. It should be noted that the packaging box does not use an NFC chip tag solution because: the carbide end mill inside the packaging has a metal shielding effect on the near-field magnetic flux of the NFC reader, which significantly increases the failure rate of reading the NFC tag embedded in the packaging box. The reliability in the industrial field cannot meet the requirements of outbound verification. In addition, the NFC tag is attached to the outer wall of the packaging, and there is no strong physical correlation between the model information stored in it and the actual items inside the packaging. This cannot prevent incorrect pairing after the tag is separated from the tool, and it essentially has the same limitations as barcode scanning.

[0030] Before implementing the requisition verification process, a benchmark database needs to be established. Because there is a nonlinear correspondence between the electromagnetic integral characteristics and the actual physical mass of the internal metal parts due to the skin effect, and this correspondence varies depending on the tool model, a regression model needs to be established for each model. For at least a predetermined number of genuine tool samples corresponding to the model identification, data pairs of spatial domain electromagnetic integral quantity and actual metal mass are collected, covering multiple representative rolling attitude angles. The actual metal mass is obtained by independently weighing the net mass of the tools inside the packaging using a precision electronic balance. For example, 40 genuine samples of M12 four-flute carbide end mills are taken, and data are collected once at each of the five attitude angles of 0°, 30°, 45°, 60°, and 90°, resulting in 200 data pairs. A multinomial regression is performed with the spatial domain electromagnetic integral quantity as the input variable and the actual metal mass as the output variable to obtain the regression model. The regression model includes model coefficients and the corresponding upper bound of the model residuals.

[0031] The above-mentioned settings of multiple representative tumbling posture angles are intended to cover various posture distributions caused by the initial tumbling of the packaging box P when it is pushed out of the slot and its collision with the inner wall of the guide channel during the fall. This ensures that the training set of the regression model covers the range of posture changes that may occur in actual operation, and the statistical fluctuations of tumbling posture have been taken into consideration in the model coefficients and the upper bound of the model residuals.

[0032] The following section uses the M12 four-flute flat end mill as an example to detail the sample set preparation and regression model training steps in the benchmark database establishment process. First, prepare the sample set: Obtain 40 genuine M12 end mills from the supplier, and weigh each end mill individually using a precision electronic balance to determine its net weight (i.e., the mass of the metal body after removing packaging), thus obtaining the actual metal mass. 40 samples The weight distribution ranged from 98.6g to 112.4g, with a mean of 105.2g and a standard deviation of 2.8g. Each tool was repackaged into its original box at five representative roll angles: 0°, 30°, 45°, 60°, and 90°. Spatial domain electromagnetic integral values ​​were collected one by one through the detection channel of the industrial tool management equipment. Each attitude angle was sampled three times and the average was taken, resulting in a total of 200 sets. Data pairs. After removing 5 sets of samples where speed estimation failed due to packaging box jamming or abnormal photoelectric node obstruction, 195 valid data pairs were obtained. The distribution range is from 84.3 kHz·mm to 99.7 kHz·mm.

[0033] The regression model uses a second-order polynomial structure, i.e. To avoid overfitting, a second-order polynomial form was chosen instead of a higher-order one. In different embodiments, the regression model could also use a polynomial structure of other orders. The coefficients were solved using the least squares method on 195 sets of training data, yielding... ,Right now After regression, the predicted residuals for each training sample are calculated. The 95th percentile of the absolute value of the residuals for all training samples is used as the upper bound δ of the model residuals. For the M12 milling cutter, δ = 3.1g. An additional 20 reserved samples are used as the validation set. Substituting the values ​​into the regression model yielded the predicted metal mass. The absolute values ​​of the prediction residuals for all 20 validation samples were below δ, with the largest residual in the validation set being 2.6g. All prediction residuals in the validation set were within acceptable limits. The regression coefficients... The upper bound δ of the model residuals is written into the regression model field of the M12 model entry in the benchmark database.

[0034] In some embodiments, the determination of the judgment threshold is also completed during the sample set acquisition phase. The threshold is determined from each sample in the 195 valid samples. The ratio is calculated by taking the mean μ and standard deviation σ of all samples. As the allowable range for the characteristic ratio, the allowable range for the characteristic ratio of the M12 end mill is [1.45, 1.72]. The packaging quality tolerance range is determined by the mean and tolerance range of the net weight samples. Twenty empty boxes were weighed individually; the mean net weight was 42g, and the tolerance was ±12g, resulting in a packaging quality tolerance range of [30g, 54g]. The allowable range for axial through-length is determined from 195 valid samples. The measured distribution range was determined using the 2% to 98th percentile of the sample distribution as the boundary. The allowable axial passage length range for the M12 end mill is [138mm, 168mm]. It should be understood that different types of tools need to establish their own regression models and judgment thresholds according to the above process, and store them in the benchmark database with the model identifier as the index.

[0035] Figure 1 A flowchart illustrating a non-unpacking verification method for an industrial tool management device according to some embodiments of this application is shown. A requisition request is obtained, which includes the model identifier of the target tool. For example, an operator initiates a milling cutter requisition request by entering the model identifier M12 through an interactive interface. Based on this model identifier, the corresponding verification benchmark set is retrieved from a benchmark database. The verification benchmark set includes the regression model coefficients, eigenvalue tolerance range [1.45, 1.72], packaging quality tolerance range [35g, 52g], and axial passage length tolerance range [140mm, 165mm] for the M12 milling cutter.

[0036] The multi-axis coil includes a first-axis coil 41, a second-axis coil 42, and a third-axis coil 43, which are orthogonally fixed to the outer wall of the non-metallic guide channel 20, and a first-LC oscillation circuit, a second-LC oscillation circuit, and a third-LC oscillation circuit, which are electrically connected to the first-axis coil 41, the second-axis coil 42, and the third-axis coil 43, respectively. Since the resonant frequency of the LC oscillation circuit drifts with changes in ambient temperature, using absolute frequency as the criterion would cause the judgment threshold to fail due to daily temperature fluctuations. Therefore, temperature drift needs to be eliminated differentially before each verification. With the non-metallic guide channel 20 in an unloaded state, the three reference frequencies output by the first, second, and third LC oscillation circuits are collected. For example, at a current temperature of 28°C, the three reference frequencies are collected as 850.2kHz, 851.5kHz, and 849.8kHz, respectively. These three values ​​are stored in a reference database for subsequent differential calculations.

[0037] In some embodiments, the three reference frequencies are re-acquired at preset calibration time intervals, and the corresponding stored values ​​in the reference database are replaced with the re-acquired reference frequencies. For example, an empty-field reference calibration is performed every 2 hours to ensure that the reference frequencies correspond to the current temperature environment. If the acquisition variance of any reference frequency exceeds a preset reference variance threshold, the system stops accepting requests and outputs a calibration anomaly alarm. For example, when a large motor near the non-metallic guide channel 20 starts and generates strong magnetic field interference, the acquisition variance of a certain reference frequency may suddenly increase to 5 times the normal value, causing the system to shut down and output an alarm to avoid misjudgments under interference conditions.

[0038] After being pushed out of the slot, the packaging box falls through the detection area via the non-metallic guide channel 20. The real-time output frequencies of the first, second, and third LC oscillation circuits, as well as the timestamp sequences of occlusion at multiple photoelectric nodes, are collected during the packaging box's passage through the detection area. The corresponding reference frequencies are subtracted from the three real-time output frequencies to obtain the three frequency offset curves. Differential operation uses temperature drift as a common mode to cancel it out, and the three frequency offset curves only reflect the net inductive response caused by the metal parts inside the packaging box.

[0039] As mentioned earlier, when packaging box P is pushed out of the slot, it gains an initial tumbling angular velocity because one end detaches from the support first. Combined with irregular collisions with the inner wall of the guide channel during its descent, the tumbling posture of packaging box P as it passes through the multi-axis electromagnetic detection zone is random. Due to the randomness of the tumbling posture of the packaging box as it falls into the non-metallic guide channel 20, the induction cross-section of the single-axis coil is extremely small when the target tool and the coil axis are nearly parallel, resulting in drastic fluctuations in the amplitude of the single-channel frequency offset curve. By performing time-synchronized three-dimensional vector synthesis of the three-channel frequency offset curves, the transient magnetoresistance modulus curve is obtained. Regardless of the three-dimensional tumbling posture of the packaging box, the projection components of the three axes are always complementary in three orthogonal directions. The synthesized modulus is independent of the posture, so that the subsequent integral features always stably correspond to the real physical volume and material of the tool.

[0040] Here, because the packaging box accelerates within the non-metallic guide channel 20, directly integrating the transient magnetoresistance modulus curve in the time domain results in a shorter time to traverse the detection area due to a faster falling speed, leading to a smaller time-domain integration area. This can cause the system to misinterpret it as a small cutting tool. Therefore, it's necessary to combine the traversal speed to map the time-domain integration to the spatial domain integration, eliminating the influence of falling speed differences. The occlusion timestamp sequence of four photoelectric nodes is obtained. These four nodes are arranged at equal intervals along the axial direction of the detection area, with a fixed spacing of ΔL = 50mm between adjacent nodes. For example, the occlusion timestamps of the four nodes during the packaging box's traversal are as follows: Based on this, the instantaneous velocities of the three sub-intervals are calculated: the velocity of interval 1. The speed of interval 2 Speed ​​in interval 3 The velocities of the three sub-intervals increase sequentially, reflecting the accelerated descent of the packaging box under the influence of gravity.

[0041] It is important to note that the instantaneous velocity of the aforementioned partition is determined by the ratio of the fixed physical distance between adjacent photoelectric nodes to the corresponding blocking time interval. This calculation process does not depend on the instantaneous attitude angle of the packaging box P when it passes through the photoelectric nodes. Therefore, the tumbling motion of the packaging box P due to its initial flip does not affect the accuracy of the calculation of the partition's instantaneous velocity. Similarly, the spatial domain electromagnetic integral obtained by multiplying the time domain integrals of each segment by the corresponding partition's instantaneous velocity and then summing them is also unaffected by the tumbling state.

[0042] Next, the transient magnetoresistance magnitude curve is divided into three segments according to the timestamp, and the time-domain integrals for each segment are as follows: The spatial domain electromagnetic integral is obtained by multiplying the integral in each time domain by the velocity in the corresponding sub-interval and then summing the results. .

[0043] The above describes one implementation method for extracting the instantaneous velocity of a partition using a photoelectric node array and converting the time-domain integral into a spatial-domain electromagnetic integral. In other embodiments, a microwave radar flow velocity sensor can also be used to obtain the continuous instantaneous velocity curve of the packaging box. Unlike the segmented velocity measurement of the photoelectric node, the microwave radar flow velocity sensor directly outputs a continuous velocity variable that changes with time based on the Doppler effect. After integrating this continuous velocity variable with the transient magnetoresistance magnitude curve point by point, the time-domain signal can be continuously mapped to the spatial-domain physical length feature without the need for segmented approximation. The microwave radar flow velocity sensor can also eliminate the influence of the falling velocity change on the integral feature, and therefore it also belongs to the implementation method of converting the transient magnetoresistance magnitude curve into a spatial-domain electromagnetic integral.

[0044] In some embodiments, the transient magnetoresistance modulus curve is further envelope-filtered to determine the geometric continuity of the internal metal parts. Here, because the carbide end mill has helical grooves along its axial direction, its cross-sectional area changes periodically. If the number of peaks is directly counted on the original M(t) curve, a genuine end mill will also exhibit multiple amplitude peaks during its rotation and descent, triggering false judgments. Using twice the time required for the maximum helical groove spacing of the target tool to pass through the detection area as the sliding window length, a sliding maximum value envelope filter is applied to the transient magnetoresistance modulus curve to obtain the envelope curve. The envelope filter smooths the high-frequency geometric texture caused by the helical grooves into a low-frequency envelope, so that subsequent peak judgment only perceives the overall outline of the internal metal body rather than local geometric details. In response to the envelope curve having a single peak, the internal metal parts of the packaging box are determined to be a single continuous body. For example, a packaging box containing a complete, intact M12 end mill will have an envelope curve with a smooth main peak. In response to the presence of multiple independent peaks in the envelope curve with an interval longer than the sliding window length, an abnormal signal is output. For example, if there are three loose blades inside, the envelope curve will show three independent peaks, with the interval between each peak far exceeding the window length. The system will directly output an abnormal signal to intercept them. Small fluctuations in the envelope curve with amplitudes less than 20% of the main peak amplitude are not counted in the independent peak count to eliminate interference from the geometric texture residuals of the spiral groove.

[0045] Based on the instantaneous velocity of each of the three sub-intervals and the corresponding occlusion time interval, the path components of each sub-interval are summed to obtain the axial passing length of the packaging box within the detection area. For example, the axial passing length... The axial passage length falls within the allowable range [140mm, 165mm] corresponding to the M12 type end mill. As a third physical dimension independent of the electromagnetic integral characteristic and the weighing characteristic, the axial passage length directly constrains the effective projected size of the object inside the packaging box in the flow direction.

[0046] The total mass of the packaging box at the weighing position is collected. Since the packaging box falls from different heights through the grids, the impact kinetic energy varies, resulting in elastic oscillations of varying durations upon landing on the receiving tray 50. If a value is directly acquired after a fixed delay, the oscillations may not have decayed by the time the drop from some grids is significant, leading to substantial reading errors. Therefore, a variance stability window criterion is used to ensure that the packaging box has nearly come to a stop at the time of acquisition. It is important to note that the landing posture and impact direction of the packaging box P as it rolls to the receiving tray 50 are random, resulting in varying decay times for the elastic oscillations. The rolling variance window criterion uses whether the real-time calculated rolling variance is below a preset stability threshold as the acquisition condition, rather than a fixed delay. Therefore, regardless of the landing posture of the packaging box P, the system only extracts the total mass reading after the oscillations have fully decayed, and the randomness of the landing posture does not affect the final accuracy of the total mass measurement.

[0047] The system collects the continuous output of the weighing sensor 60 located at the weighing position and calculates the rolling variance of the continuous output within a preset duration. If the rolling variance falls below a preset stability threshold, the average of the continuous outputs within the preset duration is used as the total mass. For example, the preset duration is 500ms, and the stability threshold... The rolling variance decreased approximately 820ms after the packaging box reached the weighing position. If the value is below the stability threshold, the average value of the continuous output over 500ms, which is 145.3g, is taken as the total mass. If the rolling variance of the packaged box remains below a preset stability threshold for more than a preset timeout period after it reaches the weighing position of the load cell 60, the system outputs a timeout exception signal and terminates the current verification process. For example, if the packaged box gets stuck at the end of the non-metallic guide channel 20 and does not fall completely, and the stability condition is not met after more than 3 seconds, the system outputs a timeout exception signal.

[0048] The above describes one implementation method for obtaining the total mass using the variance stability window criterion. In other embodiments, a fixed-delay sampling method can also be used: after the packaging box reaches the weighing position, the current reading is directly collected as the total mass after a preset delay. Unlike the variance window criterion, the fixed-delay method does not rely on real-time variance calculation and is suitable for scenarios where the drop across each compartment is consistent and the impact oscillation decay time is stable and predictable. The fixed-delay method can also obtain the total mass after the packaging box tends to come to a stop, and therefore it is also an implementation method for collecting the total mass.

[0049] Substituting the spatial domain electromagnetic integral into the regression model in the validation benchmark set yields the predicted metal quality. For example, using... Substituting into the regression model of the M12 end mill The predicted metal mass is obtained. Here, due to the continuous presence of cutting fluid mist in the machining workshop, the paper packaging boxes are in a high-humidity environment for extended periods, causing their weight to fluctuate randomly depending on the degree of moisture absorption. If the total mass is directly compared to a fixed expected value for the metal mass, the humidity fluctuations in the packaging mass will interfere with the judgment of the tool body. By extracting the predicted metal mass from the total mass, the estimated packaging mass can be obtained. This achieves data decoupling between packaging materials and the internal metal entities.

[0050] Based on the spatial domain electromagnetic integral, axial passage length, total mass, and estimated packaging quality, and compared against the judgment thresholds in the verification benchmark set, a verification pass signal or an abnormal signal is output. The judgment thresholds include the allowable range of characteristic ratio, the packaging quality tolerance range, and the allowable range of axial passage length.

[0051] Scenario 1 (Passed Normally): Feature Ratio The value falls within the allowable range of [1.45, 1.72]; estimate the packaging weight. ,fall into Within the tolerance range; axial through length ,fall into Within the permitted range. If all three conditions are met, a verification pass signal is output, and the equipment unlocks the hatch for operators to retrieve materials.

[0052] Scenario 2 (Internal metal density mismatch): If the packaging contains an equivalent mass of ordinary steel end mill (density approximately 7.8 g / cm³),... 3 It has a density lower than that of cemented carbide, which is 14.5 g / cm³. 3 For the same mass, a steel end mill has a larger volume and generates a larger inductive cross-sectional area when it crosses the detection area. The value is too high at 165.4 kHz·mm. At this point, the characteristic ratio... If the value falls outside the allowed range of [1.45, 1.72], an abnormal signal will be output.

[0053] Scenario 3 (High Mass Due to Moist Packaging): If the paper packaging of the same genuine M12 end mill absorbs a lot of water in a high-humidity environment, the total mass increases to 162g. However, the spatial domain electromagnetic integral is not affected by the packaging mass. The predicted metal quality remains at 91.8 kHz·mm, obtained from regression inversion. Still 104.9g. Estimating packaging weight. If the weight exceeds the upper limit of the packaging quality tolerance range by 52g, an abnormal signal is output. Therefore, by decoupling the data of the packaging from the metal entity, moisture in the packaging no longer causes genuine knives to be mistakenly judged as acceptable; instead, it is accurately identified as a packaging abnormality for management personnel to review and handle.

[0054] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0055] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A non-unpacking verification method for industrial tool management equipment, characterized in that, include: S1. Obtain a requisition request, wherein the requisition request includes the model identifier of the target tool; Based on the model identifier, retrieve the verification benchmark set corresponding to the model identifier from the benchmark database. The verification benchmark set includes a regression model and a judgment threshold. S2. Collect the multi-channel frequency offset curves generated by the multi-axis coils in the detection area of ​​the packaging box, as well as the occlusion timestamp sequence of multiple photoelectric nodes; determine the crossing speed of the packaging box in the detection area based on the occlusion timestamp sequence; perform vector synthesis on the multi-channel frequency offset curves to obtain the transient magnetoresistance modulus curve; combine the crossing speed to convert the transient magnetoresistance modulus curve into a spatial domain electromagnetic integral quantity; S3. Based on the occlusion timestamp sequence, obtain the axial length of the packaging box; S4. Collect the total mass of the packaging box at the weighing position; wherein the weighing position is set in a preset drop endpoint area; S5. In response to successfully obtaining the total mass, substitute the spatial domain electromagnetic integral into the regression model in the verification benchmark set to obtain the predicted metal mass; based on the difference between the total mass and the predicted metal mass, obtain the estimated packaging mass; S6. Based on the spatial domain electromagnetic integral, the axial passage length, the total mass, and the estimated packaging mass, and by comparing with the judgment threshold in the verification benchmark set, output a verification pass signal or an abnormal signal; The multi-axis coil includes a first axis coil, a second axis coil, and a third axis coil, which are orthogonally fixed to the outer wall of the non-metallic guide channel in pairs, as well as a first LC oscillation circuit, a second LC oscillation circuit, and a third LC oscillation circuit, which are electrically connected to the first axis coil, the second axis coil, and the third axis coil, respectively. The non-metallic guide channel is an insulated hollow tubular channel that runs through the detection area. The detection area is the inner cavity space of the non-metallic guide channel. The packaging box falls through the inner cavity of the non-metallic guide channel and passes through the detection area. The process of collecting the multi-channel frequency offset curves generated by the multi-axis coil during the process of the packaging box passing through the detection area, as well as the occlusion timestamp sequence of multiple photoelectric nodes, includes the following sub-steps: S21. When the non-metallic guide channel is in an unloaded state, collect the three reference frequencies output by the first LC oscillation circuit, the second LC oscillation circuit, and the third LC oscillation circuit respectively; S22. Collect the real-time output frequencies of the first LC oscillation circuit, the second LC oscillation circuit, and the third LC oscillation circuit during the process of the packaging box passing through the detection area, as well as the occlusion timestamp sequence of the multiple photoelectric nodes; S23. Subtract the corresponding reference frequency from the real-time output frequency of the first LC oscillation circuit, the second LC oscillation circuit and the third LC oscillation circuit respectively to obtain the multi-channel frequency offset curve; S24. Perform time-synchronized root mean square calculation on the frequency offset of each path in the multi-path frequency offset curve to obtain the transient magnetoresistance magnitude curve; determine the crossing speed according to the blocking timestamp sequence; and convert the transient magnetoresistance magnitude curve into the spatial domain electromagnetic integral quantity in combination with the crossing speed. In step S2, determining the crossing speed of the packaging box through the detection area based on the occlusion timestamp sequence and converting the transient magnetoresistance modulus curve into a spatial domain electromagnetic integral includes the following sub-steps: Obtain the occlusion timestamp sequence of N photoelectric nodes, where N is an integer not less than 4; Based on the fixed spacing between adjacent photoelectric nodes and the corresponding occlusion time interval, the instantaneous velocity of each of the N-1 sub-intervals is obtained; The transient magnetoresistance modulus curve is divided into N-1 segments according to the occlusion timestamp sequence. The time domain integral of each segment is multiplied by the instantaneous velocity of the corresponding sub-interval and then summed to obtain the spatial domain electromagnetic integral. The fixed spacing between adjacent photoelectric nodes is no greater than one-third of the axial coverage length of the detection area.

2. The non-unpacking verification method for industrial tool management equipment according to claim 1, characterized in that, S21 further includes: The three reference frequencies are re-acquired at preset calibration time intervals, and the corresponding stored values ​​in the reference database are replaced with the re-acquired three reference frequencies. If the sampling variance of any reference frequency exceeds the preset reference variance threshold, stop accepting requests and output a calibration anomaly alarm.

3. The non-unpacking verification method for industrial tool management equipment according to claim 1, characterized in that, In step S3, the axial through length is obtained through the following steps: Based on the instantaneous velocity of each of the N-1 sub-intervals and the corresponding occlusion time interval, the path component of each sub-interval is obtained; The axial passage length of the packaging box within the detection area is obtained by summing the path components of each of the N-1 sub-intervals.

4. The non-unpacking verification method for industrial tool management equipment according to claim 1, characterized in that, S2 further includes: Using twice the time required for the maximum helical groove spacing of the target tool corresponding to the model identifier to pass through the detection area as the sliding window length, the maximum sliding value envelope filter is applied to the transient magnetoresistance modulus curve to obtain the envelope curve; In response to the envelope curve having a single peak, it is determined that the internal metal parts of the packaging box are a single continuous body; In response to the presence of multiple independent peaks in the envelope curve with an interval length greater than the length of the sliding window, the abnormal signal is output; wherein fluctuations in the envelope curve with amplitudes lower than a preset proportion of the amplitude of the main peak of the envelope curve are not included in the count of independent peaks.

5. The non-unpacking verification method for industrial tool management equipment according to claim 1, characterized in that, S4 includes the following sub-steps: S41. Collect the continuous output of the weighing sensor located at the weighing position, and calculate the rolling variance of the continuous output within a preset duration; S42. In response to the rolling variance being lower than a preset stability threshold, the average of the continuous outputs within the preset duration is used as the total mass; S43. In response to the fact that the rolling variance is still not lower than the preset stability threshold after a preset timeout period has elapsed since the packaging box reached the weighing position of the weighing sensor, output a timeout exception signal and terminate the current verification process.

6. The non-unpacking verification method for industrial tool management equipment according to claim 1, characterized in that, In step S6, the determination threshold includes the allowable range of feature ratio, the allowable range of packaging quality tolerance, and the allowable range of axial passing length. S6 includes the following sub-steps: In response to the fact that the ratio of the total mass to the electromagnetic integral in the spatial domain falls within the characteristic ratio allowable range, the estimated packaging mass falls within the packaging mass tolerance range, and the axial passing length falls within the axial passing length allowable range, the verification pass signal is output. If any condition is not met, the abnormal signal is output.

7. The non-unpacking verification method for industrial tool management equipment according to claim 1, characterized in that, Prior to S1, the method further includes the step of establishing the benchmark database, which includes the following sub-steps: For no less than a preset number of genuine cutting tool samples corresponding to the model identifier, data pairs of spatial domain electromagnetic integral quantity and actual metal mass are collected under the condition of covering multiple representative rolling posture angles. Using the spatial domain electromagnetic integral as the input variable and the actual metal mass as the output variable, a multinomial regression is performed to obtain the regression model, which includes model coefficients and corresponding upper bounds of model residuals.

Images