Material sorting detection system and method based on intelligent sensor

Abstract

The application discloses a material sorting detection system and method based on an intelligent sensor, and belongs to the technical field of intelligent sensor material sorting detection. The method collects vibration waveform sequences generated when the material to be sorted flows through a detection area through an array piezoelectric sensor, generates a residence time parameter according to the time interval between adjacent peak points in the vibration waveform sequences, uses the residence time parameter to segment and intercept the vibration waveform sequences, and obtains independent vibration beats corresponding to each material particle. The cumulative energy value of the waveform rising edge and the curvature radius sequence of the falling edge in the independent vibration beat are extracted, the cumulative energy value is compared with a preset elastic collision energy threshold to generate a first sorting trigger signal, the coordinate position of the value mutation in the curvature radius sequence is mapped as a material surface hardness identification bit, and the belonging sorting channel of the current material particle is determined according to the logical combination state of the first sorting trigger signal and the hardness identification bit.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensor material sorting and detection technology, specifically to a material sorting and detection system and method based on intelligent sensors. Background Technology

[0002] During material sorting and inspection, array-type piezoelectric sensors can collect the vibration waveforms generated when the material to be sorted flows through the detection area, and identify the material particles based on parameters such as the amplitude, frequency, or duration of the vibration waveform. This type of detection method is used in material sorting to distinguish material particles with different collision characteristics or surface conditions.

[0003] For multiple material particles continuously passing through the detection area, the sensor output waveform is formed by the superposition of the vibration responses of multiple particles, resulting in blurred waveform boundaries between adjacent particles. Existing processing methods typically use fixed time windows or overall peak distribution for segmentation, without considering the actual residence time of individual material particles within the detection area. This leads to a situation where a single analysis window may contain vibrational components from particles preceding and following each other. When particle spacing is uneven or local signal amplitudes are similar, fixed window segmentation is prone to misalignment, causing subsequent extraction of energy and morphological features to be interfered with by signals from adjacent materials, affecting the independent determination of individual particles. To suppress noise, some solutions perform weighted synthesis of the outputs from multiple sensing units, but the synthesized waveform still retains the continuous superposition characteristic.

[0004] In existing material sorting methods, the sorting criteria are mostly based on single features such as waveform amplitude threshold, peak frequency, or overall energy. These single features have limited ability to characterize material collision characteristics and surface hardness differences. When material particles have similar elastic collision strengths but different surface hardnesses, or similar surface hardness but different elastic collision strengths, a single feature cannot provide effective differentiation. Especially in the same vibration waveform, the rising and falling edges contain different collision contact stage information; relying solely on overall statistics is insufficient to simultaneously reflect collision energy and surface hardness, easily leading to unclear channel assignment.

[0005] To reduce channel misjudgment in continuous material sorting, it is necessary to separate the independent vibration signal segment corresponding to a single material particle from the continuous vibration waveform, and simultaneously extract waveform features that can characterize the strength of elastic collision and surface hardness from the independent vibration signal segment for channel determination. Summary of the Invention

[0006] The purpose of this invention is to provide a material sorting and detection system and method based on intelligent sensors. The system collects vibration waveform sequences using an array of piezoelectric sensors, separates the independent vibration beats of each material particle based on residence time parameters, extracts the cumulative energy value of the rising edge and the curvature radius sequence of the falling edge from each independent vibration beat, and logically combines the first sorting trigger signal generated by comparing the cumulative energy value threshold with the hardness flag mapped by the curvature radius abrupt change to determine the sorting channel to which the current material particle belongs. This reduces the impact of particle signal aliasing in continuous vibration waveforms on sorting judgment and improves sorting reliability under differences in collision energy and surface hardness.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a material sorting and detection method based on intelligent sensors, comprising: acquiring a vibration waveform sequence generated when the material to be sorted flows through a detection area using an array of piezoelectric sensors; generating a residence time parameter of the material in the detection area based on the time interval between adjacent peak points in the vibration waveform sequence; segmenting the vibration waveform sequence using the residence time parameter to obtain independent vibration beats corresponding to each material particle; extracting the cumulative energy value of the rising edge and the curvature radius sequence of the falling edge of the waveform in the independent vibration beat; comparing the cumulative energy value with a preset elastic collision energy threshold to generate a first sorting trigger signal; mapping the coordinate positions of numerical abrupt changes in the curvature radius sequence to material surface hardness identifiers; and determining the sorting channel to which the current material particle belongs based on the logical combination state of the first sorting trigger signal and the hardness identifier. This solution effectively segments the continuous vibration waveforms of different material particles using the residence time parameter and achieves multi-channel sorting through the logical combination of the energy threshold and the hardness identifier, thereby improving the accuracy and efficiency of material sorting.

[0008] As a technical solution of the present invention, when acquiring vibration waveform sequences using an array of piezoelectric sensors, the analog voltage output of each sensing unit in the array of piezoelectric sensors within a continuous time window is converted into a digital quantity according to a fixed sampling frequency, forming a multi-channel parallel time sequence. Using the time sequence of the sensing unit at the center of the array of piezoelectric sensors as a reference, the phase delay of the time sequences of the remaining sensing units relative to the reference is calculated. The output values ​​of multiple sensing units whose phase delays fall within a preset angle range are weighted and superimposed to obtain a synthetic vibration waveform. The moments when the output values ​​of all sensing units exceed a start-up threshold are marked as valid acquisition start points, and waveform segments of a fixed duration are extracted from these valid acquisition start points as the vibration waveform sequence. This preferred solution enhances the effective vibration signal and suppresses asynchronous interference through phase delay filtering and weighted superposition.

[0009] As a technical solution of the present invention, when generating the residence time parameter of the material in the detection area, all local maxima points in the vibration waveform sequence are traversed, and the timestamp corresponding to each local maxima point is recorded; local maxima points with adjacent timestamps are paired, and the difference between the timestamp of the latter local maxima point and the timestamp of the former local maxima point in each pair is calculated as the initial interval value; intervals with values ​​less than the minimum material throughput threshold are counted and merged into adjacent intervals until all intervals are greater than or equal to the minimum material throughput threshold, and the merged intervals are used as the separation intervals between material particles; intervals with values ​​less than the upper limit of contact time are marked as effective residence intervals, and the arithmetic mean of all effective residence intervals is used as the residence time parameter. The residence time parameter obtained in this way can more accurately reflect the actual contact duration between material particles and the detection area, and improve the rationality of subsequent waveform segmentation.

[0010] As a technical solution of the present invention, when segmenting the vibration waveform sequence using the residence time parameter, the starting sampling point of the vibration waveform sequence is taken as the current truncation starting point. From the current truncation starting point, the search proceeds backward to the sampling point where the amplitude first exceeds the material trigger threshold, and this sampling point is marked as the beat start boundary. From the beat start boundary, the sampling point is shifted backward by a number of sampling points equal to the residence time parameter, and the reached sampling point is marked as the beat end boundary. All sampling points between the beat start boundary and the beat end boundary are extracted to form an independent vibration beat. The beat end boundary is used as the new current truncation starting point, and the search and extraction steps are repeated until the entire vibration waveform sequence has been traversed. This solution can accurately segment continuous vibration waveforms according to the actual passage time of material particles, reducing the impact of waveform aliasing between particles on subsequent feature extraction.

[0011] As a technical solution of the present invention, when extracting the cumulative energy value of the rising edge and the curvature radius sequence of the falling edge in the independent vibration beat, the peak sampling point where the amplitude reaches its maximum value is located within the independent vibration beat, and all sampling points before the peak sampling point are divided into rising edge intervals; the square value of the amplitude of each sampling point within the rising edge interval is calculated, and the sum of all square values ​​is multiplied by the time interval between adjacent sampling points to obtain the cumulative energy value; all sampling points after the peak sampling point are divided into falling edge intervals; three consecutive sampling points are taken sequentially within the falling edge interval to construct a triangle, the circumcircle radius of the triangle is calculated, and the circumcircle radius is used as the curvature radius of the middle sampling point among the three sampling points; the curvature radius sequence is obtained by traversing the falling edge interval. This preferred solution simultaneously extracts waveform energy features and falling edge curvature features, providing multi-dimensional basis for subsequent sorting and judgment.

[0012] As a technical solution of the present invention, when generating the first sorting trigger signal, the initial velocity value and nominal mass value of the current material particles before entering the detection area are obtained, and a reference kinetic energy is calculated based on the initial velocity value and the nominal mass value; the value of the elastic collision energy threshold is determined based on the ratio of the reference kinetic energy to the residence time parameter; when the cumulative energy value is greater than the elastic collision energy threshold, the first sorting trigger signal is set to a high level; when the cumulative energy value is less than or equal to the elastic collision energy threshold, the first sorting trigger signal is set to a low level. This solution dynamically adjusts the threshold according to the material kinetic energy and residence time, avoiding misjudgment caused by a fixed threshold under different operating conditions.

[0013] As a technical solution of the present invention, when mapping the coordinate positions of numerical abrupt changes in the radius of curvature sequence to material surface hardness identifiers, the absolute value of the difference between two adjacent radius of curvature values ​​in the radius of curvature sequence is calculated to obtain a difference sequence; the coordinate positions corresponding to the differences in the difference sequence that are greater than the abrupt change judgment threshold are marked as abrupt change coordinates; the distribution density of the abrupt change coordinates in the radius of curvature sequence is statistically analyzed, and the surface roughness level of the material is determined based on the distribution density; the surface roughness level is matched with the hardness level in a preset hardness mapping table, and the binary code corresponding to the matched hardness level is used as the hardness identifier. This solution converts the curvature abrupt change feature of the falling edge of the waveform into a hardness identifier, realizing non-contact identification of material hardness.

[0014] As a technical solution of the present invention, when determining the sorting channel to which the current material particle belongs based on the logical combination state of the first sorting trigger signal and the hardness identifier bit, when the first sorting trigger signal is at a high level and the value corresponding to the hardness identifier bit is in the first hardness range, the current material particle is assigned to the first sorting channel; when the first sorting trigger signal is at a high level and the value corresponding to the hardness identifier bit is in the second hardness range, the current material particle is assigned to the second sorting channel; when the first sorting trigger signal is at a low level and the value corresponding to the hardness identifier bit is in the first hardness range, the current material particle is assigned to the third sorting channel; and when the first sorting trigger signal is at a low level and the value corresponding to the hardness identifier bit is in the second hardness range, the current material particle is assigned to the fourth sorting channel. This solution achieves four-channel sorting through the combined logic of two features, improving sorting resolution and adaptability to operating conditions.

[0015] As a technical solution of the present invention, during the weighted superposition, the weighting coefficient of the output value of each sensing unit is determined according to the phase delay between the sensing unit and the center position sensing unit, and the weighting coefficient is proportional to the cosine value of the phase delay. This allows the signals of each channel to be effectively enhanced after phase alignment, suppressing asynchronous interference.

[0016] This invention also provides a material sorting and detection system based on intelligent sensors, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the aforementioned material sorting and detection method based on intelligent sensors. This system can be applied to industrial material sorting scenarios, utilizing vibration waveforms collected by array-type piezoelectric sensors to complete material feature extraction and logical sorting, achieving automated and highly efficient sorting operations.

[0017] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0018] After the array-type piezoelectric sensor collects the vibration waveform sequence generated when the material to be sorted flows through the detection area, the initial interval value is determined by the time interval between adjacent peak points. Intervals smaller than the minimum material passage time threshold are merged, and intervals with values ​​smaller than the upper limit of contact time are taken as effective retention intervals. The retention time parameter is obtained by the arithmetic mean of the effective retention intervals. When segmenting the vibration waveform sequence using the retention time parameter, the sampling point where the amplitude first exceeds the material trigger threshold is searched from the current segmentation starting point as the beat start boundary. The number of sampling points corresponding to the retention time parameter is pushed forward as the beat end boundary. The sampling points between the two boundaries constitute an independent vibration beat corresponding to a single material particle. Since the segmentation boundary is determined according to the actual retention time of the material, it no longer depends on a fixed time window or the overall peak distribution. Even if the spacing of multiple material particles passing through is uneven or the local amplitude is similar, they can still be separated according to their respective passage time through the detection area. The vibration response of adjacent particles will not enter the same analysis window. The extracted rising and falling edge features all come from the collision process of the same material particle, reducing the interference of signal aliasing on subsequent energy calculation and hardness identification extraction.

[0019] The cumulative energy value of the rising edge and the curvature radius sequence of the falling edge are extracted within an independent vibration cycle. The rising edge interval consists of sampling points before the peak sampling point with the largest amplitude within the cycle, and the cumulative energy value is obtained by multiplying the square of the amplitude of each sampling point by the time interval between adjacent sampling points and summing the results. The falling edge interval consists of sampling points after the peak sampling point, and the curvature radius sequence is obtained by constructing a triangle by three consecutive sampling points and calculating the radius of the circumcircle. The cumulative energy value is compared with a preset elastic collision energy threshold to generate the first sorting trigger signal. The coordinates of adjacent differences in the curvature radius sequence that are greater than the abrupt change judgment threshold are marked as abrupt change coordinates. The unevenness level of the material surface is determined according to the distribution density of the abrupt change coordinates and a hardness identifier is matched. Finally, the sorting channel is determined by the logical combination state of the first sorting trigger signal and the hardness identifier. In the same independent vibration cycle, the cumulative energy value of the rising edge reflects the strength of the elastic collision when the material collides with the detection area, and the sudden change in the radius of curvature of the falling edge maps the surface hardness information of the material. By combining the two features in a logical way for sorting, it is possible to distinguish between material particles with similar collision energy but different surface hardness and material particles with similar surface hardness but different collision energy. This expands the sorting dimensions that cannot be covered by a single amplitude or frequency criterion and reduces channel misjudgment caused by the overlap of a single feature. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0021] Figure 1 This is a flowchart of a material sorting and detection method based on intelligent sensors;

[0022] Figure 2 This is a flowchart of the method for generating material sorting trigger signals;

[0023] Figure 3 This is a scatter plot comparing the cumulative energy value with the elastic collision energy threshold;

[0024] Figure 4 This is a distribution diagram of the material particle classification and sorting channels. Detailed Implementation

[0025] 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, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] See Figure 1 This invention provides a material sorting and detection method based on intelligent sensors, comprising: acquiring a vibration waveform sequence generated when the material to be sorted flows through a detection area using an array of piezoelectric sensors; generating a residence time parameter of the material in the detection area based on the time interval between adjacent peak points in the vibration waveform sequence; segmenting the vibration waveform sequence using the residence time parameter to obtain independent vibration beats corresponding to each material particle; extracting the cumulative energy value of the rising edge and the curvature radius sequence of the falling edge of the waveform in the independent vibration beat; comparing the cumulative energy value with a preset elastic collision energy threshold to generate a first sorting trigger signal; mapping the coordinate positions of numerical abrupt changes in the curvature radius sequence to material surface hardness identifiers; and determining the sorting channel to which the current material particle belongs based on the logical combination state of the first sorting trigger signal and the hardness identifier.

[0027] In practical implementation, the array-type piezoelectric sensor comprises multiple sensing units arranged in a two-dimensional array. The analog voltage output of each sensing unit within a continuous time window is transmitted to an analog-to-digital converter (ADC). The ADC converts the analog voltage output of each sensing unit into a digital signal at a fixed sampling frequency. The fixed sampling frequency is determined based on the highest frequency component of the vibration signal output by the array-type piezoelectric sensor and is greater than twice the highest frequency component. The digital signals obtained from each sensing unit are arranged in the sampling time sequence to form a time series, and multiple sensing units form multiple parallel time series.

[0028] Using the time series of the sensing unit at the center of the array piezoelectric sensor as a reference, calculate the phase delay of the time series of each sensing unit in the array piezoelectric sensor (excluding the center sensing unit) relative to the reference. When calculating the phase delay, the [missing information - likely a specific unit or parameter] is used... The time series of the first sensing unit is cross-correlated with the time series of the center position sensing unit. The sampling point offset corresponding to the maximum value in the cross-correlation result is multiplied by the sampling period to obtain the first... The time delay between each sensing unit and the sensing unit at the center position. The time delay of the first sensing unit relative to the center position sensing unit is multiplied by the angular frequency of the vibration signal to obtain the first... The phase delay of each sensing unit relative to the center sensing unit. This refers to the index of the sensing unit number in the array piezoelectric sensor, excluding the center position sensing unit. Take a positive integer.

[0029] Preset angle range set to The preset angle range is set to... The basis is: when the phase delay is in Within the specified range, the cosine of the phase delay is non-negative, and the corresponding weighting coefficient is also non-negative, preventing anti-phase cancellation during weighted superposition. The output values ​​of multiple sensing units with phase delays within the preset angle range are weighted and superimposed to obtain the synthesized vibration waveform. During the weighted superposition process, the output value of each sensing unit with a phase delay within the preset angle range is multiplied by the weighting coefficient corresponding to that sensing unit. All multiplication results are summed, and the summation is used as the amplitude of the synthesized vibration waveform at the corresponding sampling time.

[0030] When weighted and superimposed, the first The weighting coefficient of the output value of the first sensing unit is based on the first... The phase delay between each sensing unit and the central position sensing unit is determined. The weighting coefficient of the output value of the first sensing unit and the weighting coefficient of the first sensing unit. The phase delay of each sensing unit relative to the center sensing unit is proportional to the cosine value, satisfying the following relationship:

[0031]

[0032] in, For the first Weighting coefficients for the output values ​​of each sensing unit; For the first The phase delay of each sensing unit relative to the central sensing unit; This is the index of the sensor unit whose phase delay is within a preset angle range. Take a positive integer. The weighting coefficient of the output value of the i-th sensing unit is directly taken from the j-th... When the cosine of the phase delay of each sensing unit relative to the center sensing unit is taken, the weighting coefficient corresponding to the sensing unit with a phase delay of zero is 1, and the phase delay is... or The weighting coefficient corresponding to the sensing unit is 0.

[0033] The moment when the output value of all sensing units exceeds the trigger threshold is marked as the valid acquisition start point. The trigger threshold is determined based on the root mean square (RMS) value of the output noise of all sensing units in the array piezoelectric sensor when no material passes through. The trigger threshold is set to 3 times the RMS value of the output noise when no material passes through. The basis for setting the trigger threshold to 3 times the RMS value of the output noise when no material passes through is that it is greater than the upper limit of the normal fluctuation of the output noise amplitude when no material passes through. A waveform segment of fixed duration is extracted from the valid acquisition start point, and the extracted waveform segment is used as the vibration waveform sequence. The fixed duration is determined based on the contact duration of the material particles passing through the detection area, and the fixed duration covers the time range from the material particles contacting the detection area to leaving the detection area.

[0034] In practice, all local maxima in the vibration waveform sequence are traversed. A local maximum is a sampling point in the vibration waveform sequence whose amplitude is greater than both the amplitude of the preceding and following sampling points. The timestamp corresponding to each local maximum is recorded; the timestamp is the sampling point number of the local maximum in the vibration waveform sequence.

[0035] Pair up local maxima with adjacent timestamps. Local maxima with adjacent timestamps refer to two local maxima that are adjacent in position in a sequence of local maxima arranged in chronological order. The difference between the timestamp of the latter local maximum and the timestamp of the former local maximum in each pair is used as the initial interval value.

[0036] The initial interval values ​​are counted for those less than the minimum material throughput time threshold. The minimum material throughput time threshold is determined based on the nominal size of the particles to be sorted and the lower limit of their movement speed within the detection area. It is set as the ratio of the nominal size of the particles along the direction of movement to the lower limit of their movement speed. Initial interval values ​​less than the minimum material throughput time threshold are merged into adjacent initial interval values. The merging operation involves adding each initial interval value less than the minimum material throughput time threshold to its adjacent initial interval value; the sum is used as the merged initial interval value. This counting and merging operation is repeated until all initial interval values ​​are greater than or equal to the minimum material throughput time threshold. The merged initial interval value is then used as the separation interval between the material particles.

[0037] Separation intervals with values ​​less than the upper limit of contact time are marked as effective retention intervals. The upper limit of contact time is determined based on the upper limit of the elastic collision duration between the particles to be sorted and the detection area. The upper limit of contact time is greater than the minimum material throughput threshold. The arithmetic mean of all effective retention intervals is used as the retention time parameter. The retention time parameter is calculated according to the following formula:

[0038]

[0039] in, This refers to the residence time parameter; The number of effective dwell intervals. Take a positive integer; For the first The value of an effective dwell interval, For the sequence number index of the effective dwell interval, Take from 1 to A positive integer. When the number of valid stay intervals... When the value is zero, the residence time parameter is set to the arithmetic mean of the upper limit of the contact time and the minimum material throughput threshold.

[0040] When segmenting a vibration waveform sequence using the residence time parameter, the initial sampling point of the vibration waveform sequence is taken as the current segmentation starting point. Starting from the current segmentation starting point, the search proceeds backward to the sampling point where the amplitude first exceeds the material trigger threshold. The material trigger threshold is determined based on the root mean square (RMS) value of the noise amplitude in the vibration waveform sequence when no material passes through. The material trigger threshold is set to four times the RMS value of the noise amplitude in the vibration waveform sequence when no material passes through. The basis for setting the material trigger threshold to four times the RMS value of the noise amplitude in the vibration waveform sequence when no material passes through is that it is greater than the upper limit of the random fluctuation of the noise amplitude when no material passes through. The sampling point where the amplitude first exceeds the material trigger threshold is marked as the beat start boundary.

[0041] Starting from the beat start boundary, shift backward by a number of sampling points equal to the dwell time parameter. The number of sampling points is the floor function of the product of the dwell time parameter and the sampling frequency. Mark the sampling point reached after shifting backward as the beat end boundary. Extract all sampling points between the beat start boundary and the beat end boundary to form an independent vibration beat. An independent vibration beat includes the beat start boundary sampling point, the beat end boundary sampling point, and all sampling points between the beat start boundary and the beat end boundary.

[0042] The cycle termination boundary is used as the new current interception starting point. Starting from this new starting point, a search is performed backwards for sampling points whose amplitude first exceeds the material trigger threshold; these points are marked as the new cycle start boundary. From this new start boundary, sampling points are moved backwards by a number of points equal to the residence time parameter; these points are marked as the new cycle termination boundary. All sampling points between the new cycle start boundary and the new cycle termination boundary are extracted to form the next independent vibration cycle. This search and extraction process is repeated until the entire vibration waveform sequence has been traversed. After traversing the entire vibration waveform sequence, if the number of remaining sampling points after the last cycle termination boundary is less than the number of points equal to the residence time parameter, the remaining sampling points are discarded.

[0043] By using the vibration waveform sequence and beat division shown in the figure, the processor can accurately extract independent vibration beats. Subsequently, energy accumulation and radius of curvature calculation can be performed based on the vibration characteristics of each beat to achieve the sorting and detection of material particles. The process of vibration waveform sequence extraction and beat boundary determination reflected in the figure is completely consistent with the method description of residence time parameter generation and independent vibration beat segmentation in the above embodiment.

[0044] In specific implementation, please refer to Figure 2 Within an independent vibration beat, the peak sampling point where the amplitude reaches its maximum value is located. The peak sampling point refers to the sampling point with the largest amplitude among all sampling points contained in the independent vibration beat. When multiple sampling points with equal maximum amplitude exist within the independent vibration beat, the sampling point with the earliest timestamp is taken as the peak sampling point. All sampling points before the peak sampling point are divided into rising edge intervals, which include all sampling points from the starting boundary sampling point of the independent vibration beat to the sampling point preceding the peak sampling point. All sampling points after the peak sampling point are divided into falling edge intervals, which include all sampling points from the sampling point following the peak sampling point to the ending boundary sampling point of the independent vibration beat.

[0045] The calculation process for the cumulative energy value is as follows: Calculate the square of the amplitude at each sampling point within the rising edge interval. Sum the squares of the amplitudes at all sampling points within the rising edge interval. The time interval between adjacent sampling points is the sampling period, which is equal to the reciprocal of the sampling frequency. Multiply the sum of all squares by the time interval between adjacent sampling points; the product is the cumulative energy value.

[0046] The process of obtaining the descent edge curvature radius sequence is as follows: Within the descent edge interval, three consecutive sampling points are taken sequentially according to their sampling time. A triangle is constructed using the sampling time of the three consecutive sampling points as the x-axis and the amplitude of the three consecutive sampling points as the y-axis. The three sampling points are, in order, the previous sampling point, the middle sampling point, and the next sampling point. The lengths of the three sides of the triangle are calculated: the first side is the Euclidean distance between the previous and middle sampling points, the second side is the Euclidean distance between the middle and next sampling points, and the third side is the Euclidean distance between the previous and next sampling points. The area of ​​the triangle is calculated using the lengths of its three sides. The area of ​​the triangle is calculated using Heron's formula, where the semi-perimeter is half the sum of the lengths of the three sides. The circumradius of the triangle is equal to the product of the lengths of the three sides divided by four times the area of ​​the triangle. The calculated circumradius is taken as the curvature radius of the middle sampling point among the three sampling points. Within the falling edge interval, starting from the initial position, calculate the radius of curvature every three consecutive sampling points. Then, move one sampling point forward and repeat the process of constructing triangles and calculating the circumcircle radius until there are fewer than three sampling points within the falling edge interval. Arrange the radii of curvature corresponding to each intermediate sampling point within the falling edge interval in chronological order of sampling time to obtain the radius of curvature sequence.

[0047] The process for determining the elastic collision energy threshold is as follows: Obtain the initial velocity and nominal mass of the current material particle before it enters the detection area. The initial velocity is measured by a velocity sensor positioned upstream of the detection area, and the nominal mass is measured by a weighing sensor positioned at the same location as the velocity sensor. Half of the product of the square of the initial velocity and the nominal mass is taken as the reference kinetic energy, which is calculated using the following formula:

[0048]

[0049] in, As the baseline kinetic energy, This represents the nominal mass value of the current material particles. This represents the initial velocity value of the material particles before they enter the detection area. The value is output by the weighing sensor. The value is output by the speed sensor.

[0050] The elastic collision energy threshold is set as the ratio of the reference kinetic energy to the dwell time parameter. The dwell time parameter is obtained by the method in the above embodiment. When the accumulated energy value is greater than the elastic collision energy threshold, the first sorting trigger signal is set to a high level. When the accumulated energy value is less than or equal to the elastic collision energy threshold, the first sorting trigger signal is set to a low level. The first sorting trigger signal is a digital signal, with a high level corresponding to a logic value of 1 and a low level corresponding to a logic value of 0.

[0051] See Figure 3 In the figure, the horizontal axis represents the particle number, ranging from 0 to 500, indicating the sequentially collected particle numbers; the vertical axis represents the relative unit of the cumulative energy value, reflecting the cumulative energy value of each particle corresponding to an independent vibration beat. Two different markers are used in the figure to distinguish the relationship between the cumulative energy value and the elastic collision energy threshold: red circles represent particles with a cumulative energy value greater than the elastic collision energy threshold, and blue squares represent particles with a cumulative energy value less than or equal to the elastic collision energy threshold; the dashed line in the figure represents the elastic collision energy threshold, whose value is approximately 1.9 of the vertical axis.

[0052] As can be seen from the curve distribution in the graph, the cumulative energy values ​​are clearly divided into two groups: one group is above the dashed line, where the cumulative energy value is greater than the elastic collision energy threshold; the other group is below the dashed line, where the cumulative energy value is less than or equal to the elastic collision energy threshold. Most of the red circles are distributed between 2 and 6 on the vertical axis, showing a concentrated peak but with some fluctuations, indicating differences in energy release during elastic collisions. The blue squares are mainly concentrated between 0.5 and 2 on the vertical axis, with a smaller fluctuation range, indicating that the cumulative energy corresponding to material particles that did not reach the elastic collision energy threshold is lower.

[0053] This diagram illustrates the threshold comparison process between the cumulative energy value and a preset elastic collision energy threshold in one of the above embodiments. According to the diagram, the processor can accurately distinguish the collision energy state of material particles by comparing the cumulative energy value with the elastic collision energy threshold, thereby generating a high or low level for the first sorting trigger signal. Specifically, materials with a cumulative energy value greater than the threshold are determined to be particles with high elastic collision energy, and the trigger signal is set to a high level; materials with a cumulative energy value less than or equal to the threshold have a corresponding low level trigger signal.

[0054] In specific implementation, the absolute value of the difference between two adjacent radius of curvature values ​​in the radius of curvature sequence is calculated to obtain the difference sequence. The radius of curvature sequence is formed by arranging the radii of curvature corresponding to each intermediate sampling point within the falling edge interval in the above embodiment in chronological order of sampling time. The total number of radius of curvature values ​​in the radius of curvature sequence is denoted as . , It is a positive integer. Greater than or equal to 3. The absolute value of the difference is calculated using the following formula:

[0055]

[0056] in, The first in the difference sequence The absolute value of the difference; The first in the radius of curvature sequence One radius of curvature value; The first in the radius of curvature sequence One radius of curvature value; This is the index of the difference element in the difference sequence. Take from 1 to positive integers, Let be the total number of radius values ​​in the radius of curvature sequence. Iterate through all adjacent radius values ​​in the radius of curvature sequence, calculate the absolute value of the difference for each pair according to the formula above, and arrange all the absolute values ​​of the difference in the adjacent order of the radius of curvature sequence to obtain the difference sequence.

[0057] In some embodiments, the mutation detection threshold is determined based on the arithmetic mean of all curvature radius values ​​in the curvature radius sequence. The mutation detection threshold is set to 0.5 times the arithmetic mean of all curvature radius values ​​in the curvature radius sequence. The basis for setting the mutation detection threshold to 0.5 times the arithmetic mean of all curvature radius values ​​in the curvature radius sequence is that the normal fluctuation range of curvature radius values ​​is concentrated around the arithmetic mean, and a difference exceeding 0.5 times the arithmetic mean indicates a significant jump in the curvature radius. The coordinate positions corresponding to the differences in the difference sequence that are greater than the mutation detection threshold are marked as mutation coordinates. The first... The coordinate position corresponding to the difference value is the th in the radius of curvature sequence. The position of each radius of curvature value. The abrupt change coordinates are recorded as the sampling position number in the radius of curvature sequence.

[0058] In practice, the distribution density of abrupt change coordinates within the radius of curvature sequence is statistically analyzed. The distribution density is set as the ratio of the number of abrupt change coordinates to the total number of radius of curvature values ​​in the radius of curvature sequence. The distribution density ranges from 0 to 1. The surface roughness level of the material is determined based on the distribution density. The roughness levels include a first roughness level, a second roughness level, and a third roughness level. The first roughness level corresponds to a distribution density less than 0.2, the second roughness level corresponds to a distribution density greater than or equal to 0.2 and less than 0.5, and the third roughness level corresponds to a distribution density greater than or equal to 0.5. The first roughness level indicates a low degree of surface roughness, the second roughness level indicates a moderate degree of surface roughness, and the third roughness level indicates a high degree of surface roughness.

[0059] Optionally, a preset hardness mapping table contains the correspondence between surface roughness levels and hardness levels. The preset hardness mapping table stores the correspondence between a first surface roughness level and a first hardness level, a second surface roughness level and a second hardness level, and a third surface roughness level and a third hardness level. The surface roughness level of the material is matched with the surface roughness levels in the preset hardness mapping table, and the resulting hardness level is used as the hardness level of the current material particles. The binary code corresponding to the hardness level is used as the hardness identifier. The binary code corresponding to the first hardness level is 01, the binary code corresponding to the second hardness level is 10, and the binary code corresponding to the third hardness level is 11.

[0060] In some embodiments, the value corresponding to the hardness identifier is the decimal value corresponding to the binary code of the hardness identifier. The decimal value of the binary code 01 corresponding to the first hardness level is 1, the decimal value of the binary code 10 corresponding to the second hardness level is 2, and the decimal value of the binary code 11 corresponding to the third hardness level is 3. The first hardness range is the range of decimal values ​​less than or equal to 1, and the second hardness range is the range of decimal values ​​greater than 1.

[0061] In practice, the sorting channel assigned to the current material particle is determined based on the logical combination of the first sorting trigger signal and the hardness indicator bit. When the first sorting trigger signal is high and the value corresponding to the hardness indicator bit is within the first hardness range, the current material particle is assigned to the first sorting channel. When the first sorting trigger signal is high and the value corresponding to the hardness indicator bit is within the second hardness range, the current material particle is assigned to the second sorting channel. When the first sorting trigger signal is low and the value corresponding to the hardness indicator bit is within the first hardness range, the current material particle is assigned to the third sorting channel. When the first sorting trigger signal is low and the value corresponding to the hardness indicator bit is within the second hardness range, the current material particle is assigned to the fourth sorting channel. A high level for the first sorting trigger signal corresponds to a logic value of 1, and a low level corresponds to a logic value of 0. After the assigned sorting channel for the current material particle is determined, the sorting execution mechanism guides the current material particle into the corresponding material collection area according to the assigned sorting channel.

[0062] In practical implementation, the material sorting and detection system based on intelligent sensors includes a memory, a processor, and a computer program stored in the memory and running on the processor. The memory and processor are connected via a data bus. The processor reads computer program instructions from the memory via the data bus and executes the computer program instructions. The memory is divided into a program storage area and a data storage area. The program storage area is used to store the computer program, and the data storage area is used to store intermediate data and output data generated by the processor during the execution of the computer program.

[0063] In practical implementation, the processor executes the computer program to realize the vibration waveform sequence acquisition process. The processor receives the analog voltage output of each sensing unit in the array piezoelectric sensor and converts it into a digital quantity after analog-to-digital conversion. The processor synchronously acquires multiple digital quantities at a fixed sampling frequency. Taking the time series of the sensing unit at the center position in the array piezoelectric sensor as a reference, the processor calculates the phase delay of the time series of each sensing unit in the array piezoelectric sensor (excluding the sensing unit at the center position) relative to the time series of the sensing unit at the center position. The processor weighted and superimposed the output values ​​of multiple sensing units whose phase delay is within a preset angle range to obtain a synthetic vibration waveform. The processor marks the moment when the output value of all sensing units exceeds the start threshold as the valid acquisition start point. The processor extracts a waveform segment of fixed duration from the valid acquisition start point as the vibration waveform sequence.

[0064] In practical implementation, the processor executes the computer program to generate the residence time parameter. The processor iterates through all local maxima in the vibration waveform sequence, recording the timestamp corresponding to each local maxima. The processor pairs adjacent local maxima with each timestamp, calculating the difference between the timestamp of the latter local maxima and the timestamp of the former local maxima in each pair. The processor uses this difference as the initial interval value. The processor counts the initial interval values ​​whose values ​​are less than the minimum material throughput threshold. The processor merges these initial interval values ​​into adjacent initial interval values. The processor repeats the counting and merging operations until all initial interval values ​​are greater than or equal to the minimum material throughput threshold. The processor uses the merged initial interval values ​​as the separation intervals between material particles. The processor marks separation intervals whose values ​​are less than the upper limit of the contact time as effective residence intervals. The processor uses the arithmetic mean of all effective residence intervals as the residence time parameter.

[0065] In practice, the processor executes the computer program to implement an independent vibration beat extraction process. The processor takes the starting sampling point of the vibration waveform sequence as the current extraction starting point; starting from the current extraction starting point, the processor searches backward for the sampling point where the amplitude first exceeds the material trigger threshold, and marks the sampling point where the amplitude first exceeds the material trigger threshold as the beat start boundary; starting from the beat start boundary, the processor moves backward for a number of sampling points equal to the residence time parameter, and marks the sampling point reached after the move as the beat end boundary; the processor extracts all sampling points between the beat start boundary and the beat end boundary, and the processor combines all the extracted sampling points into an independent vibration beat; the processor uses the beat end boundary as the new current extraction starting point, and the processor repeats the search and extraction operations until the entire vibration waveform sequence has been traversed.

[0066] In practical implementation, the processor executes the computer program to extract the cumulative energy value and radius of curvature sequence. The processor locates the peak sampling point where the amplitude reaches its maximum value within an independent vibration beat. The processor divides all sampling points before the peak sampling point into rising edge intervals. The processor calculates the square of the amplitude of each sampling point within the rising edge interval, sums the squares of all sampling points within the rising edge interval, and multiplies the sum by the time interval between adjacent sampling points. The processor uses the product as the cumulative energy value. The processor divides all sampling points after the peak sampling point into falling edge intervals. Within the falling edge interval, the processor sequentially selects three consecutive sampling points according to the sampling time order to construct a triangle. The processor calculates the radius of the circumcircle of the triangle and uses this radius as the radius of curvature of the middle sampling point among the three sampling points. The processor iterates through the falling edge intervals to obtain the radius of curvature sequence.

[0067] In practical implementation, the processor executes the computer program to generate the first sorting trigger signal. The processor acquires the initial velocity and nominal mass of the current material particles before entering the detection area. The processor uses half of the product of the square of the initial velocity and the nominal mass as the reference kinetic energy. The processor determines the value of the elastic collision energy threshold based on the ratio of the reference kinetic energy to the residence time parameter. The processor compares the cumulative energy value with the elastic collision energy threshold. When the cumulative energy value is greater than the elastic collision energy threshold, the processor sets the first sorting trigger signal to a high level. When the cumulative energy value is less than or equal to the elastic collision energy threshold, the processor sets the first sorting trigger signal to a low level.

[0068] In practical implementation, the processor executes the computer program to perform the hardness identifier mapping process. The processor calculates the absolute value of the difference between two adjacent radius values ​​in the radius of curvature sequence, and arranges all the absolute values ​​of the difference in the adjacent order of the radius of curvature sequence to obtain the difference sequence; the processor marks the coordinate positions corresponding to the differences in the difference sequence that are greater than the abrupt change judgment threshold as abrupt change coordinates; the processor calculates the distribution density of abrupt change coordinates in the radius of curvature sequence, and determines the surface roughness level of the material based on the distribution density; the processor matches the surface roughness level of the material with the hardness levels in the preset hardness mapping table, and uses the binary code corresponding to the matched hardness level as the hardness identifier bit.

[0069] In practical implementation, the processor executes the computer program to determine the assigned sorting channel. The processor determines the assigned sorting channel for the current material particle based on the logical combination of the first sorting trigger signal and the hardness identifier bit. When the first sorting trigger signal is high and the value corresponding to the hardness identifier bit is within the first hardness range, the processor assigns the current material particle to the first sorting channel. When the first sorting trigger signal is high and the value corresponding to the hardness identifier bit is within the second hardness range, the processor assigns the current material particle to the second sorting channel. When the first sorting trigger signal is low and the value corresponding to the hardness identifier bit is within the first hardness range, the processor assigns the current material particle to the third sorting channel. When the first sorting trigger signal is low and the value corresponding to the hardness identifier bit is within the second hardness range, the processor assigns the current material particle to the fourth sorting channel. The processor outputs the control signal corresponding to the assigned sorting channel of the current material particle to the sorting execution mechanism.

[0070] See Figure 4 In the graph, the horizontal axis represents the particle number of the material, ranging from 0 to 500, and the vertical axis represents the sorting channel to which the material particle belongs, with a value ranging from 1 to 4. In the legend, the scatter dots of different colors and shapes represent the first sorting channel (blue dots), the second sorting channel (red squares), the third sorting channel (green triangles), and the fourth sorting channel (orange diamonds), respectively.

[0071] The four groups of scattered points in the figure are evenly distributed along the direction of the material particle number, and the assigned channel markings are clear, showing the result of the assigned sorting channel determination process implemented by the processor in the above embodiment through a computer program. Specifically, the processor correctly classifies the material particles into the corresponding sorting channels based on the logical combination state of the first sorting trigger signal and the hardness identifier bit, which is represented by a group of discrete points with different channel numbers on the vertical axis.

[0072] The values ​​for the first sorting channel are concentrated around the vertical axis 1, showing a relatively concentrated distribution without significant fluctuations. This indicates that the corresponding material particles meet the logical conditions of a high-level first sorting trigger signal and the hardness indicator being in the first hardness range. The scatter points for the second sorting channel are concentrated around the vertical axis 2, showing a uniform and relatively stable distribution. This reflects that the corresponding material particles meet the condition of a high-level first sorting trigger signal and the hardness indicator being in the second hardness range. The scatter points for the third sorting channel are concentrated around the vertical axis 3, showing a uniform distribution. This indicates that the material particles correspond to the condition of a low-level first sorting trigger signal and the hardness indicator being in the first hardness range. The scatter points for the fourth sorting channel are concentrated around the vertical axis 4, showing a stable distribution, which meets the logical combination of a low-level first sorting trigger signal and the hardness indicator being in the second hardness range.

[0073] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A material sorting and detection method based on intelligent sensors, characterized in that, include: The vibration waveform sequence generated when the material to be sorted flows through the detection area is collected by an array of piezoelectric sensors; Based on the time interval between adjacent peak points in the vibration waveform sequence, the residence time parameter of the material in the detection area is generated; The vibration waveform sequence is segmented and truncated using the residence time parameter to obtain independent vibration beats corresponding to each material particle; Extract the cumulative energy value of the rising edge and the curvature radius sequence of the falling edge of the waveform in the independent vibration beat; The accumulated energy value is compared with a preset elastic collision energy threshold to generate a first sorting trigger signal; The coordinate positions of numerical abrupt changes in the radius of curvature sequence are mapped to material surface hardness identifier bits; Based on the logical combination state of the first sorting trigger signal and the hardness identifier, the sorting channel to which the current material particle belongs is determined.

2. The material sorting and detection method based on intelligent sensors according to claim 1, characterized in that, The vibration waveform sequence generated when the material to be sorted flows through the detection area, acquired by an array of piezoelectric sensors, includes: The analog voltage output of each sensing unit in the array piezoelectric sensor within a continuous time window is converted into a digital quantity at a fixed sampling frequency, forming a multi-channel parallel time series. Using the time series of the sensing unit located at the center of the array piezoelectric sensor as a reference, the phase delay of the time series of the remaining sensing units relative to the reference is calculated. The output values ​​of multiple sensing units whose phase delay is within a preset angle range are weighted and superimposed to obtain a synthetic vibration waveform. The moment when the output value of all sensing units exceeds the start threshold is marked as the effective acquisition start point, and a waveform segment of fixed duration is extracted from the effective acquisition start point as the vibration waveform sequence.

3. The material sorting and detection method based on intelligent sensors according to claim 2, characterized in that, The step of generating the residence time parameter of the material in the detection area based on the time interval between adjacent peak points in the vibration waveform sequence includes: Traverse all local maxima in the vibration waveform sequence and record the timestamp corresponding to each local maxima. Pair up local maxima points with adjacent timestamps, and calculate the difference between the timestamp of the latter local maximum point and the timestamp of the former local maximum point in each pair, as the initial interval value. The interval values ​​that are less than the minimum material throughput time threshold in the initial interval values ​​are counted and merged into adjacent interval values ​​until all interval values ​​are greater than or equal to the minimum material throughput time threshold. The merged interval values ​​are then used as the separation intervals between material particles. Intervals with values ​​less than the upper limit of contact time are marked as effective retention intervals, and the arithmetic mean of all effective retention intervals is used as the retention time parameter.

4. The material sorting and detection method based on intelligent sensors according to claim 3, characterized in that, The step of segmenting the vibration waveform sequence using the residence time parameter to obtain independent vibration beats corresponding to each material particle includes: Using the starting sampling point of the vibration waveform sequence as the current truncation starting point, search backwards from the current truncation starting point for sampling points where the amplitude first exceeds the material trigger threshold, and mark the sampling point as the beat start boundary; Starting from the beat start boundary, move backward by a number of sampling points equal to the dwell time parameter, and mark the sampling points reached after the move as the beat end boundary; Extract all sampling points between the start boundary and the end boundary of the beat to form an independent vibration beat; Using the beat termination boundary as the new current interception starting point, repeat the search and extraction steps until the entire vibration waveform sequence has been traversed.

5. The material sorting and detection method based on intelligent sensors according to claim 4, characterized in that, The extraction of the cumulative energy value of the rising edge and the curvature radius sequence of the falling edge of the waveform in the independent vibration beat includes: Within the independent vibration beat, locate the peak sampling point where the amplitude reaches its maximum value, and divide all sampling points before the peak sampling point into rising edge intervals; Calculate the square of the amplitude of each sampling point within the rising edge interval, sum all the squares and multiply by the time interval between adjacent sampling points to obtain the cumulative energy value; All sampling points after the peak sampling point are divided into falling edge intervals; Within the falling edge interval, three consecutive sampling points are taken to construct a triangle. The radius of the circumcircle of the triangle is calculated, and the radius of the circumcircle is used as the radius of curvature of the middle sampling point among the three sampling points. The radius of curvature sequence is obtained by traversing the falling edge interval.

6. The material sorting and detection method based on intelligent sensors according to claim 5, characterized in that, The step of comparing the accumulated energy value with a preset elastic collision energy threshold to generate a first sorting trigger signal includes: Obtain the initial velocity value and nominal mass value of the current material particles before entering the detection area, and take half of the product of the square of the initial velocity value and the nominal mass value as the reference kinetic energy; The value of the elastic collision energy threshold is determined based on the ratio of the reference kinetic energy to the dwell time parameter. When the accumulated energy value is greater than the elastic collision energy threshold, the first sorting trigger signal is set to a high level. When the cumulative energy value is less than or equal to the elastic collision energy threshold, the first sorting trigger signal is set to a low level.

7. The material sorting and detection method based on intelligent sensors according to claim 6, characterized in that, Mapping the coordinate positions of numerical abrupt changes in the radius of curvature sequence to material surface hardness identifiers includes: Calculate the absolute value of the difference between two adjacent radius of curvature values ​​in the radius of curvature sequence to obtain the difference sequence; The coordinate positions corresponding to the differences in the difference sequence that are greater than the mutation determination threshold are marked as mutation coordinates; The distribution density of the abrupt change coordinates in the radius of curvature sequence is statistically analyzed, and the unevenness level of the material surface is determined based on the distribution density. The unevenness level is matched with the hardness level in the preset hardness mapping table, and the binary code corresponding to the matched hardness level is used as the hardness identifier bit.

8. The material sorting and detection method based on intelligent sensors according to claim 7, characterized in that, The step of determining the current material particle's assigned sorting channel based on the logical combination state of the first sorting trigger signal and the hardness identifier bit includes: When the first sorting trigger signal is high and the value corresponding to the hardness identifier is in the first hardness range, the current material particles are assigned to the first sorting channel. When the first sorting trigger signal is high and the value corresponding to the hardness identifier is in the second hardness range, the current material particles are assigned to the second sorting channel. When the first sorting trigger signal is in a low-level state and the value corresponding to the hardness identifier is in the first hardness range, the current material particles are assigned to the third sorting channel. When the first sorting trigger signal is in a low-level state and the value corresponding to the hardness identifier is in the second hardness range, the current material particles are assigned to the fourth sorting channel.

9. The material sorting and detection method based on intelligent sensors according to claim 2, characterized in that, During the weighted superposition, the weighting coefficient of the output value of each sensing unit is determined based on the phase delay between the sensing unit and the sensing unit at the center position, and the weighting coefficient is proportional to the cosine value of the phase delay.

10. A material sorting and detection system based on intelligent sensors, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the material sorting and detection method based on intelligent sensors as described in any one of claims 1 to 9.

Images