Pharmaceutical equipment fault detection method and system

Abstract

The invention relates to the technical field of fault detection, and particularly provides a pharmaceutical equipment fault detection method and system.The pharmaceutical equipment fault detection method comprises the steps that in the granule medicine production process, real-time operation parameters of granule medicine production equipment are obtained, acoustic response signals at different positions of the granule medicine production equipment are collected; extracting a first acoustic propagation path feature according to all the acoustic response signals; acquiring a preset acoustic propagation path change mode according to the real-time operation parameters; when the change mode of the first acoustic propagation path is inconsistent with a preset acoustic propagation path change mode, comparing the first acoustic propagation path feature with the second acoustic propagation path feature to obtain a path feature difference; analyzing characteristic change modes of different pickup sensor pairs according to the path characteristic difference and preset equipment geometric topological structure data so as to identify an acoustic characteristic abnormal region and position a fault point; according to the method, the fault of the pharmaceutical equipment can be identified and positioned.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault detection, in particular to a pharmaceutical equipment fault detection method and system. BACKGROUND

[0002] In the granule pharmaceutical production area, the automatic production line sends dry pharmaceutical granules to the packaging station through the pneumatic conveying system. In order to ensure the operation of the production line, the prior art deploys a set of blockage detection system based on acoustic resonance principle. The system installs acoustic exciter and pickup sensor on the outer wall of the pipeline, the exciter periodically injects wideband sweep acoustic signal, and the sensor collects the pipeline vibration response signal. The central processing unit analyzes the frequency spectrum of the response signal, extracts the inherent resonance frequency distribution, and compares it with the "healthy" baseline spectrum under the condition of no blockage. The offset of the resonance frequency is used to determine whether the pipeline is blocked. Once the offset exceeds the preset threshold, the system issues a warning.

[0003] However, in the actual production environment, the blockage detection method based on acoustic resonance principle faces challenges in precise blockage positioning when facing complex pipeline networks. The pipeline system of the production line contains structures such as elbows, reducers, branch pipes and valves. These structures affect the propagation path and attenuation characteristics of sound waves in the pipeline, causing reflection, scattering and interference phenomena of the signals emitted by the acoustic exciter inside the pipeline. The signal received by the pickup sensor is the superimposed response from the entire pipeline system, making the positioning of the signal source complex.

[0004] Due to space limitations in the production site, equipment installation costs and daily maintenance convenience, the acoustic exciter and pickup sensor can usually only be installed at specific discrete positions of the pipeline system, and cannot achieve dense coverage of the entire pipeline. This makes the monitoring range covered by a single sensor limited, and the signal received by it contains information from the complex pipeline structure near and far from the sensor installation point. Therefore, when the detection system identifies the resonance frequency shift, it can only determine that there may be a blockage in "some area" of the pipeline system, but cannot indicate the specific location of the blockage. The operator needs to spend time to check each section, which is inefficient in long-distance or multi-branch pipeline systems.

[0005] In a pipeline network containing branches and loops, the propagation path of acoustic signals is multiple. Even if multiple sensors are installed at different positions, when a blockage occurs in a branch pipeline, the acoustic resonance changes caused by it may propagate through multiple paths to different sensors. The signal characteristics received by these sensors may be similar, or due to signal attenuation and interference, it is difficult to distinguish from these signals which branch or specific location the blockage occurs. This leads to a decrease in the accuracy of the system in positioning the blockage point in complex networks.

[0006] During the operation of the production line, there can be changes in the material conveying flow, adjustments in the air flow pressure, or temporary stoppage / activation of different pipe sections, etc. These operations themselves can cause changes in the internal medium state of the pipe, thereby affecting the propagation of acoustic signals and the resonance characteristics. These changes can produce a frequency shift similar to a slight blockage on the acoustic spectrum, increasing the complexity of positioning. When the system receives a resonance frequency shift signal, it is difficult to distinguish whether it is a local change caused by a blockage or a change in acoustic characteristics caused by a change in the operating state, thereby causing positioning errors or ambiguous positioning and increasing the false alarm rate.

[0007] In some cases, the pipe system can not have a single blockage, but rather slight material attachment or accumulation at multiple different locations at the same time. Each slight blockage point will affect the acoustic resonance characteristics of the pipe, and these effects will superimpose together, making the total response signal received by the sensor complex. The existing system is difficult to identify and distinguish multiple independent blockage points from this superimposed signal, and it is even more difficult to divide the specific location of each blockage point. The system can only issue a general blockage warning and cannot provide specific location information for multiple blockage points, increasing the difficulty of fault diagnosis and elimination.

[0008] Therefore, the current detection system based on the principle of acoustic resonance in a complex pipe network faces the technical bottleneck of being unable to accurately and efficiently locate the blockage point.

[0009] There is currently no effective technical solution to the above problems. It should be noted that the above information disclosed in this part is only used to understand the background of the inventive concept, and therefore can contain information that does not constitute the prior art. SUMMARY

[0010] The purpose of the present application is to provide a pharmaceutical equipment fault detection method and system that can effectively solve the problem of the existing pharmaceutical equipment fault detection system in a complex pipe network and overcome the challenge of being unable to distinguish between changes caused by normal operation and actual faults.

[0011] In a first aspect, the present application provides a pharmaceutical equipment fault detection method, comprising the following steps: S1. During the production of granular drugs, real-time operating parameters of a granular drug production device are obtained, and an acoustic exciter is used to inject an acoustic signal and a plurality of pickup sensors are used to collect acoustic response signals at different positions of the granular drug production device; the real-time operating parameters include material conveying flow, air flow pressure, and activation / deactivation state of each branch pipe; S2. A first acoustic propagation path feature is extracted according to all acoustic response signals; S3, obtaining a preset acoustic propagation path change mode according to the real-time operation parameter, and analyzing whether the change mode of the first acoustic propagation path is consistent with the preset acoustic propagation path change mode; S4, when the change mode of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change mode, comparing the first acoustic propagation path feature with a second acoustic propagation path feature corresponding to the real-time operation parameter in a pipeline acoustic propagation path feature spectrum baseline database constructed in advance, to obtain a path feature difference; S5, analyzing feature change modes of different pairs of pickup sensors according to the path feature difference and preset device geometric topology structure data, to identify an acoustic characteristic abnormal area and locate a fault point.

[0012] In a second aspect, the present application further provides a pharmaceutical equipment fault detection system, which comprises: An information acquisition module is configured to acquire real-time operation parameters of a granular drug production equipment during a granular drug production process, and inject acoustic signals through an acoustic exciter and acquire acoustic response signals at different positions of the granular drug production equipment through a plurality of pickup sensors; the real-time operation parameters include material conveying flow, air flow pressure, and start / stop state of each branch pipeline; A path feature acquisition module is configured to extract a first acoustic propagation path feature according to all the acoustic response signals; An analysis module is configured to obtain a preset acoustic propagation path change mode according to the real-time operation parameter, and analyze whether the change mode of the first acoustic propagation path is consistent with the preset acoustic propagation path change mode; A difference acquisition module is configured to, when the change mode of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change mode, compare the first acoustic propagation path feature with a second acoustic propagation path feature corresponding to the real-time operation parameter in a pipeline acoustic propagation path feature spectrum baseline database constructed in advance, to obtain a path feature difference; A fault analysis module is configured to analyze feature change modes of different pairs of pickup sensors according to the path feature difference and preset device geometric topology structure data, to identify an acoustic characteristic abnormal area and locate a fault point.

[0013] As can be seen from the above, the pharmaceutical equipment fault detection method and system provided by the present application can realize identification and positioning of pharmaceutical equipment faults by comprehensively utilizing real-time operation parameters, multi-sensor acoustic data, and device topology structure information, so that the present application can effectively solve the problem that existing pharmaceutical equipment fault detection systems cannot locate fault points in complex pipeline networks and overcome the challenge of distinguishing between changes caused by normal operation and actual faults, thereby improving the accuracy and efficiency of fault diagnosis and reducing the time and cost of manual troubleshooting. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 A flow chart of a pharmaceutical equipment fault detection method provided in an embodiment of the present application.

[0015] Figure 2 A structural schematic diagram of a pharmaceutical equipment fault detection system provided in an embodiment of the present application.

[0016] The reference signs: 1, information acquisition module; 2, path feature acquisition module; 3, difference acquisition module; 4, fault analysis module. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0018] It should be noted that: similar reference signs and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms “first”, “second” and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0019] In a first aspect, as shown in the drawings, the present application provides a pharmaceutical equipment fault detection method, which comprises the following steps: Figure 1 In a first aspect, as shown in the drawings, the present application provides a pharmaceutical equipment fault detection method, which comprises the following steps: S1, in the granular drug production process, real-time running parameters of a granular drug production equipment are acquired, and acoustic signals are injected through an acoustic exciter and acoustic response signals at different positions of the granular drug production equipment are collected through a plurality of pickup sensors; the real-time running parameters include material conveying flow, air flow pressure and start / stop state of each branch pipeline; S2, a first acoustic propagation path feature and its change mode are extracted according to all the acoustic response signals; S3, a preset acoustic propagation path change mode is acquired according to the real-time running parameters, and whether the change mode of the first acoustic propagation path is consistent with the preset acoustic propagation path change mode is analyzed; S4, when the change mode of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change mode, comparing the first acoustic propagation path feature with the second acoustic propagation path feature corresponding to the real-time operation parameter in the pipeline acoustic propagation path feature spectrum baseline database constructed in advance to obtain a path feature difference; S5, according to the path feature difference and the preset device geometric topology structure data, analyzing the feature change mode of different pairs of pickup sensors to identify an acoustic characteristic abnormal area and locate a fault point.

[0020] The acoustic exciter refers to a device capable of injecting an acoustic signal into the interior of a pharmaceutical device (a granular medicine production device including a pipeline system), which can be a piezoelectric transducer, an electromagnetic vibrator, or a pneumatic sound source (such as an ultrasonic generator or a loudspeaker), and is mainly used to generate controllable sound waves in the pipeline in order to subsequently collect the propagation response thereof. The pickup sensor refers to a device capable of collecting acoustic response signals at different positions of the pharmaceutical device, which can be a microphone, an accelerometer, or a piezoelectric sensor (such as a MEMS microphone array or a vibration sensor), and is mainly used to obtain specific response information of the sound waves after propagation in the pipeline. The real-time operating parameter refers to key data reflecting the current working condition of the device during the production of granular medicine, which can include material conveying flow, air flow pressure, and the on / off state of each branch pipeline, and is mainly used to provide context information of the current production environment in order to distinguish between changes caused by normal operation and actual faults. The first acoustic propagation path feature refers to a set of characteristics reflecting the actual propagation state of sound waves in the pipeline, which can include the arrival time difference of acoustic signals, the distortion degree of signal waveforms, the amplitude attenuation amount, the phase shift amount, and the bandwidth and damping factor of the resonance peak, and is mainly used to characterize the propagation behavior of sound waves in the current pipeline. The change pattern of the first acoustic propagation path feature refers to the law of the first acoustic propagation path feature extracted from the acoustic response signals collected by the pickup sensor through the acoustic exciter injecting an acoustic signal during the production of granular medicine, i.e., the first acoustic propagation path feature of this embodiment is equivalent to the acoustic parameters directly extracted from the acoustic response signal at the current time, and the change pattern of the first acoustic propagation path feature of this embodiment is equivalent to the change law of the first acoustic propagation path feature about time extracted from the first acoustic propagation path features at different times. The change pattern of the first acoustic propagation path feature can be obtained by using existing trend analysis methods or fluctuation analysis methods to perform time series analysis on the first acoustic propagation path features at different times. The preset acoustic propagation path change pattern refers to the change law of the acoustic propagation path under normal operating conditions (such as flow adjustment, air flow pressure change, or pipeline state switching), which can be achieved by using a model trained based on historical data or a predefined rule set (such as a lookup table or a regression model), and is mainly used to establish a baseline of acoustic behavior under normal conditions in order to distinguish between normal fluctuations and abnormal situations. The pipeline acoustic propagation path feature spectrum baseline database refers to a database storing a set of pipeline acoustic propagation path features under different normal operating parameters, which can be a relational database or a non-relational database (such as a SQL database or a NoSQL database), and is mainly used to provide a reference of acoustic propagation path features under normal conditions for comparison with the real-time collected acoustic propagation path features.The path feature difference refers to the quantitative deviation degree between the first acoustic propagation path feature currently collected and the second acoustic propagation path feature corresponding to the normal state in the database, which can be calculated by using the Euclidean distance, cosine similarity or dynamic time warping algorithm, and is mainly used to measure the deviation degree of the current acoustic state from the normal state. The preset device geometric topology structure data refers to the structured data describing the physical layout, connection relationship and position information of each component of the pipeline system of the pharmaceutical device, which can be implemented by using a CAD model, a graph database or a topology graph, such as a pipeline three-dimensional model, a node-edge graph, and is mainly used to provide the physical context of the pipeline system, assist in analyzing the acoustic abnormal area and locating the fault point.

[0021] The core innovation of the present application is that by combining the real-time running parameter with the acoustic propagation path feature analysis, and using the preset device geometric topology structure data, the acoustic changes caused by normal operation and the actual faults are effectively distinguished, and the identification and positioning of the fault point in the complex pipeline network are realized, thereby achieving the effect of improving the fault diagnosis accuracy and efficiency.

[0022] Specifically, the method is initiated at the beginning of the granule drug production process. First, an acoustic exciter injects acoustic signals into the interior of the equipment, while multiple pickup sensors collect acoustic response signals at different positions in the equipment. During this process, the system synchronously acquires real-time operating parameters such as material conveying flow, air flow pressure, and the start / stop state of each branch pipe, which provide necessary information about the current production conditions and lay the foundation for subsequent analysis. Next, the system extracts a first acoustic propagation path feature from all collected acoustic response signals, which characterizes the actual propagation state of sound waves in the internal medium and structure of the pipeline. Subsequently, the system acquires a preset acoustic propagation path change pattern using the previously acquired real-time operating parameters, which represents the expected change pattern of the acoustic propagation path under the current normal working condition. The system compares the actual extracted change pattern of the first acoustic propagation path with the preset pattern to determine whether they are consistent, effectively identifying and filtering out acoustic characteristic fluctuations caused by normal production operations (such as flow adjustment, air pressure changes, or pipeline start / stop), thereby avoiding false positives. When the change pattern of the first acoustic propagation path is inconsistent with the preset pattern, it indicates that there is an abnormal situation (the change in the acoustic propagation path is not caused by normal production operations). At this time, the system compares the current first acoustic propagation path feature with the second acoustic propagation path feature corresponding to the current real-time operating parameters in the pre-constructed pipeline acoustic propagation path feature spectrum baseline database. Through this comparison, the system can measure the path feature difference between the current state and the corresponding normal state, providing a quantitative basis for subsequent fault analysis. Finally, the system analyzes the change pattern of the acoustic characteristics between different pairs of pickup sensors based on the obtained path feature difference and the preset equipment geometric topology structure data. By integrating this information, the system can infer the area where the acoustic characteristic anomaly occurs and locate the fault point. This method realizes the identification and positioning of pharmaceutical equipment faults through multi-dimensional data fusion and hierarchical analysis.

[0023] As a preferred embodiment, the scheme of the present application is implemented as follows: a plurality of piezoelectric acoustic exciters are installed on the outer wall of the pipeline of the granular drug production line, and a plurality of MEMS microphone arrays are deployed as pickup sensors along the pipeline and at the branch node. At the same time, sensors installed on the granular drug production line are used to collect material conveying flow, air flow pressure and the start / stop state of each branch pipeline in real time, and these real-time operating parameters are transmitted to the central control unit. After receiving the acoustic response signals collected by the microphone array, the central control unit processes the signals to extract the time difference of arrival of acoustic signals between different microphone pairs, the degree of signal waveform distortion, the amplitude attenuation, and other first acoustic propagation path characteristics. The control unit then obtains the preset acoustic propagation path change pattern under the current working condition from the pre-stored lookup table according to the real-time operating parameters, which can indicate how the acoustic signal attenuation should change when the flow is adjusted from X to Y. The control unit compares the actual extracted first acoustic propagation path change pattern with the preset pattern, and if inconsistencies are found, it is determined that there may be an anomaly. Once an anomaly is determined, the control unit accesses the pre-constructed pipeline acoustic propagation path characteristic spectrum baseline database, which stores normal acoustic propagation path characteristics under different flow, pressure and pipeline states. The control unit retrieves the corresponding second acoustic propagation path characteristics from the database according to the current real-time operating parameters, and compares them with the current first acoustic propagation path characteristics to obtain the path characteristic differences. Finally, the control unit analyzes the specific pattern of these path characteristic differences between different microphone pairs in combination with the preset pipeline three-dimensional CAD model (device geometric topology data) to identify the specific pipeline segment with abnormal acoustic characteristics and locate the blockage point. For example, if the microphone pairs upstream and downstream of a certain pipeline segment exhibit abnormal time difference of arrival and amplitude attenuation, while other areas are normal, the system can accurately infer that the acoustic characteristic abnormality area is concentrated in branch A, not in branch B or other positions of the main pipeline, according to the CAD model.

[0024] Through the above scheme, the present application can effectively solve the problem that the existing pharmaceutical equipment fault detection system cannot locate the fault point in a complex pipeline network, and overcome the challenge of distinguishing between changes caused by normal operation and actual faults. The present application realizes the identification and localization of pharmaceutical equipment faults by comprehensively utilizing real-time operating parameters, multi-sensor acoustic data and device topology information, thereby improving the accuracy and efficiency of fault diagnosis and reducing the time and cost of manual troubleshooting.

[0025] In some preferred embodiments, the first acoustic propagation path feature and the second acoustic propagation path feature each include a time-of-arrival difference, a signal waveform distortion degree, an amplitude attenuation amount, a phase shift amount, and a bandwidth and a damping factor of a resonance peak of the acoustic signal between different pairs of pickup sensors. The time-of-arrival difference refers to a difference in time required for the acoustic signal to propagate between different pairs of pickup sensors, which can be achieved using high-precision timestamp synchronization technology combined with signal cross-correlation analysis. The signal waveform distortion degree refers to a degree to which the original waveform of the acoustic signal deviates from an ideal state during propagation, which can be quantified using signal processing methods such as Fourier transform, wavelet analysis, or empirical mode decomposition. The amplitude attenuation amount refers to a magnitude of reduction in energy or intensity of the acoustic signal during propagation, which can be achieved by measuring a ratio of peak or root mean square amplitudes of the received signal to the transmitted signal. The phase shift amount refers to a deviation of the phase of the acoustic signal from a reference signal or an expected phase during propagation, which can be achieved using phase demodulation techniques or Fourier transform-based phase spectrum analysis. The bandwidth and the damping factor of the resonance peak refer to a width of a resonance frequency response curve generated by the piping system under acoustic excitation and a degree of energy dissipation of the system, which can be extracted and quantified using spectral analysis, modal analysis, or system identification techniques.

[0026] In some preferred embodiments, step S3 includes: S31, performing multi-scale decomposition on the variation pattern of the first acoustic propagation path, combining real-time operating parameters, identifying and suppressing transient noise and non-blocking interference caused by particle collisions and airflow pulsations to obtain an enhanced acoustic propagation path variation pattern; S32, obtaining a preset acoustic propagation path variation pattern according to real-time operating parameters; S33, obtaining a pattern similarity according to the enhanced acoustic propagation path variation pattern and the preset acoustic propagation path variation pattern; S34, analyzing whether the pattern similarity is greater than a preset pattern similarity threshold, if yes, determining that the variation pattern of the first acoustic propagation path is consistent with the preset acoustic propagation path variation pattern, if not, determining that the variation pattern of the first acoustic propagation path is inconsistent with the preset acoustic propagation path variation pattern.

[0027] Among them, multi-scale decomposition is a signal processing technology, which aims to decompose complex acoustic signals into components at different time scales and frequency ranges to help reveal the hidden patterns and structures in the signals and provide a basis for subsequent noise identification and suppression. Multi-scale decomposition can be achieved by using wavelet transform or empirical mode decomposition method. Since the changes of real-time operating parameters will affect the characteristics of particle collision and airflow pulsation noise. For example, when the material conveying flow increases, the frequency and intensity of particle collision may increase, causing the energy distribution of transient noise to change, so this embodiment uses real-time operating parameters to assist in noise identification to more accurately separate and remove these non-fault-related disturbances from the original acoustic propagation path change pattern. The identification and suppression of noise can be achieved by using adaptive filtering, Wiener filtering or machine learning-based noise suppression algorithm, for example, training a deep learning model that can output enhanced acoustic propagation path change pattern after noise suppression according to the input decomposed acoustic components and real-time operating parameters. This embodiment can effectively improve the purity of the acoustic propagation path change pattern by suppressing transient noise and non-blocking disturbances caused by particle collision and airflow pulsation, so as to provide a high-quality data basis for subsequent pattern comparison. The acquisition of the preset acoustic propagation path change pattern can be based on a pre-established database or prediction model, for example, a database containing normal operation patterns under different material conveying flow, airflow pressure and each branch pipe enabled / disabled state is constructed, when the system obtains the current real-time operating parameters, the corresponding preset pattern is queried from the database or calculated by interpolation, or the current real-time operating parameters are input into the machine learning model to make the model output the expected preset acoustic propagation path change pattern. This embodiment can ensure that the reference pattern used for comparison matches the current device operating state, so as to distinguish between acoustic changes caused by normal operation adjustment and abnormalities caused by potential faults. The pattern similarity is an index for quantifying the matching degree between the enhanced acoustic propagation path change pattern and the preset acoustic propagation path change pattern. The pattern similarity can be calculated based on statistical methods (such as correlation coefficient) or distance metrics (such as Euclidean distance). The pattern similarity reflects the closeness of the two patterns. It should be understood that since the enhanced pattern has removed the interference noise, and the preset pattern corresponds to the current operating parameters, the calculated pattern similarity can more truly reflect the matching degree between the current device acoustic characteristics and the expected characteristics in the normal operating state. The preset pattern similarity threshold is a reference value for judging pattern consistency. The threshold can be a fixed value, or it can be dynamically adjusted according to the real-time operating parameters to adapt to different production conditions and noise levels, thereby improving the accuracy and robustness of the judgment. For example, under certain operating parameters, the acoustic characteristics of the device may have a larger normal fluctuation range, in which case the threshold can be appropriately reduced to avoid false positives.

[0028] The application significantly improves the ability to distinguish acoustic changes caused by normal operation from actual faults in a complex production environment by introducing a mechanism for noise suppression and enhancement mode analysis in the analysis of the acoustic propagation path change mode, thereby effectively reducing the false positive rate. Specifically, after obtaining the first acoustic propagation path change mode, it is first subjected to multi-scale decomposition, which decomposes complex acoustic signals into multiple acoustic components at different time scales and frequency ranges, so that the details and macro features of the signal are clearly presented. On this basis, combined with the current real-time operating parameters, the system can accurately identify and suppress the transient noise and non-blocking interference caused by particle collision and airflow pulsation inherent in the production process to obtain an enhanced acoustic propagation path change mode that is more pure and better reflects the acoustic characteristics of the pipeline. At the same time, the system dynamically obtains the preset acoustic propagation path change mode according to the current real-time operating parameters, thereby improving the accuracy and applicability of the comparison. Subsequently, the enhanced acoustic propagation path change mode obtained is compared with the dynamically obtained preset acoustic propagation path change mode to obtain the mode similarity between the two, which can truly reflect the matching degree between the current pipeline acoustic characteristics and the normal operation mode and avoid false dissimilarity caused by noise. Finally, by analyzing whether the mode similarity is greater than the preset mode similarity threshold, the system can make a final judgment. Since the comparison is based on the enhanced mode after noise suppression, the reliability of the judgment result is greatly improved, effectively avoiding false positives caused by inherent noise and non-blocking interference in the production process, so that the system can more accurately identify real abnormal situations. Therefore, this embodiment can significantly improve the recognition accuracy of pipeline acoustic abnormalities (especially in actual production environments with complex noise interference) based on the original fault detection method, effectively distinguishing acoustic changes caused by normal operation from actual faults, thereby providing a more reliable basis for subsequent fault location.

[0029] In some preferred embodiments, step S31 comprises: S311, multi-scale decomposition is performed on the first acoustic propagation path change mode to obtain acoustic components at different time scales and frequency ranges; S312, obtaining the particle characteristics of the current production of the granular drug production equipment; the particle characteristics include the particle type and the particle size distribution; S313, obtaining the particle collision noise characteristics and the airflow pulsation noise characteristics according to the particle characteristics and the real-time operating parameters; S313, identifying the part of the acoustic components corresponding to the transient noise and non-blocking interference caused by particle collision and airflow pulsation according to the particle collision noise characteristics and the airflow pulsation noise characteristics; S314, a suppression processing is performed on a part of the acoustic component corresponding to the transient noise caused by particle collision and airflow pulsation and non-blocking interference, and an enhanced acoustic propagation path change mode is generated according to the acoustic component after the suppression processing.

[0030] The acoustic component refers to the acoustic signal component obtained after multi-scale decomposition, which has independent physical meaning in a specific time scale and frequency range. The particle characteristics refer to the physical properties of the granular drug, which can be obtained by using an online particle size analyzer, an image recognition system or laboratory sampling detection. These properties affect the acoustic characteristics generated when the particles move in the pipeline. Specifically, the particle type refers to the classification attributes of the granular drug, such as chemical composition, density and hardness. The particle size distribution refers to the proportion of particles of different sizes in the granular drug, which can be represented by average particle size, particle size range or distribution curve. The particle collision noise feature refers to the unique attributes of the acoustic signal generated by the collision of particles on the inner wall of the pipeline or between particles, such as specific frequency components, energy distribution, transient shock waveform or repetition rate. The airflow pulsation noise feature refers to the acoustic signal attributes caused by the turbulent flow, vortex or pressure fluctuation generated by the airflow during the flow in the pipeline, such as broadband noise spectrum, specific frequency harmonic or periodic fluctuation. The transient noise refers to a noise signal with short duration, concentrated energy and sudden appearance, which is usually related to the violent collision of a single or a small number of particles. The non-blocking interference refers to a signal that is not caused by pipeline blockage but may affect the acoustic propagation mode, such as acoustic changes caused by normal material transportation, airflow fluctuation or equipment vibration. The suppression processing refers to reducing or eliminating the influence of specific noise components on the overall signal by signal processing methods, which can be achieved by using adaptive filtering, wavelet threshold denoising, spectral subtraction or machine learning-based noise separation algorithm. The enhanced acoustic propagation path change mode refers to the acoustic mode after noise recognition and suppression processing, which can more accurately reflect the acoustic propagation characteristics inside the pipeline and effectively remove the non-blocking interference.

[0031] To more accurately identify and suppress the transient noise and non-clog interference caused by particle collision and airflow pulsation, the application proposes a refined processing flow. First, the change pattern of the first acoustic propagation path is subjected to multi-scale decomposition, so that the unique acoustic characteristics of noise of different sources (such as particle collision and airflow pulsation) can be revealed in specific components. On this basis, the system further acquires the particle characteristics of the current production of the particle drug production equipment. Since different types and sizes of particles will produce different collision acoustic characteristics when moving in the pipeline, the particle type and particle size distribution are the key inputs for understanding and predicting particle collision noise. At the same time, combined with real-time operating parameters (such as material conveying flow and airflow pressure), the system can more accurately obtain the particle collision noise characteristics and airflow pulsation noise characteristics under the current working condition. Subsequently, using the acquired particle collision noise characteristics and airflow pulsation noise characteristics, the system identifies the acoustic components obtained by previous decomposition to accurately find out the part of the acoustic component corresponding to the transient noise and non-clog interference caused by particle collision and airflow pulsation. By comparing the actual acoustic component with the pre-modeled noise characteristics, the acoustic changes caused by normal operation can be effectively distinguished from potential clogging signals and false positives can be avoided. Finally, the identified acoustic components corresponding to the transient noise and non-clog interference are suppressed to effectively remove these interference information and ensure that the reconstructed enhanced acoustic propagation path change pattern is more pure. Through this series of steps, the real acoustic propagation pattern related to clogging is stripped from the original acoustic signal, thereby providing more accurate and reliable basis for subsequent pattern comparison and fault judgment. This refined noise identification and suppression method enables the system to more accurately determine whether the change pattern of the first acoustic propagation path is consistent with the preset acoustic propagation path change pattern in a complex production environment, significantly reducing the false positive rate or false negative rate caused by noise due to normal operation, thereby improving the accuracy and reliability of the entire fault detection method.

[0032] In a specific embodiment, in order to achieve accurate suppression of transient noise and non-blocking interference in the first acoustic propagation path variation mode, the following operations can be performed. First, the collected first acoustic propagation path variation mode signal is subjected to multi-scale decomposition. At the same time, the system can obtain the particle characteristics of the currently produced granules from the production control system, for example, by querying the type of the current batch of granules and its particle size distribution data through the material management system. These data are input into a pre-trained noise feature model, which generates particle collision noise features and airflow pulsation noise features based on the input particle characteristics and the obtained real-time operating parameters. The model can be a machine learning model (such as a support vector machine or a neural network) that is pre-trained under different particle characteristics and operating parameters to predict or generate the spectral shape, energy distribution, and time-domain waveform features of particle collision and airflow pulsation noise under the current working condition. Subsequently, using these dynamically obtained noise features, the acoustic components obtained by wavelet packet decomposition are identified, for example, by using a correlation analysis or pattern matching method to compare the features of each acoustic component with the predicted particle collision noise features and airflow pulsation noise features. If the similarity of a certain acoustic component to these noise features exceeds a pre-set threshold, it is determined that the component contains transient noise or non-blocking interference caused by particle collision or airflow pulsation. Finally, the identified acoustic components containing noise are suppressed, for example, by using an adaptive noise cancellation algorithm to reduce or set to zero their amplitude. After processing, all processed acoustic components are recombined by wavelet packet synthesis to reconstruct the enhanced acoustic propagation path variation mode. The mode obtained in this way will significantly reduce the interference caused by normal production activities, thus more accurately reflecting the true acoustic state inside the pipeline.

[0033] In some preferred embodiments, step S34 comprises: S341, obtaining conveying flow fluctuation information and airflow pressure fluctuation information according to real-time operating parameters, and obtaining particle characteristics of the granule produced by the granule drug production equipment; S342, obtaining a pre-set mode similarity threshold according to the conveying flow fluctuation information, the airflow pressure fluctuation information, and the particle characteristics; S343, analyzing whether the mode similarity is greater than the pre-set mode similarity threshold, if yes, determining that the first acoustic propagation path variation mode is consistent with the pre-set acoustic propagation path variation mode, if not, determining that the first acoustic propagation path variation mode is inconsistent with the pre-set acoustic propagation path variation mode.

[0034] The conveying flow fluctuation information refers to dynamic data of the change of the flow of the material over time during conveying in the pipeline, which can be obtained by using a flow sensor, a mass flowmeter or a flow calculation method based on pressure difference. The gas flow pressure fluctuation information refers to dynamic data of the change of the gas flow pressure in the pipeline over time, which can be obtained by using a high-precision pressure sensor or a differential pressure sensor. The preset mode similarity threshold refers to a dynamic reference value for judging whether the change mode of the first acoustic propagation path is consistent with the preset acoustic propagation path change mode, which can be dynamically determined according to real-time operation parameters, conveying flow fluctuation information, gas flow pressure fluctuation information and particle characteristics through a pre-established mathematical model, a lookup table or a machine learning algorithm.

[0035] The scheme can effectively improve the accuracy and adaptability of fault detection because it introduces a dynamic perception and adaptive adjustment mechanism for real-time operating conditions. First, by obtaining the conveying flow fluctuation information, airflow pressure fluctuation information, and particle characteristics of the current production of the granular drug production equipment, the system can fully grasp the key environmental factors affecting the propagation of acoustic signals. These real-time parameters are the direct driving factors of the changes in the propagation characteristics of acoustic signals in the pipeline. For example, changes in flow and pressure will change the density and flow rate of the medium in the pipeline, thereby affecting the sound speed and attenuation; the type and particle size distribution of the particles will affect the collision frequency and energy of the particles with the pipe wall, producing different background noise and acoustic responses. It is precisely because of the accurate acquisition of these real-time parameters that the subsequent judgment can be based on a full consideration of the current operating conditions. On this basis, the system dynamically acquires the preset mode similarity threshold according to the real-time conveying flow fluctuation information, airflow pressure fluctuation information, and particle characteristics, which is in sharp contrast to the traditional method of using a fixed threshold. Through this dynamic adjustment, the standard for judging whether the acoustic mode is consistent is no longer rigid, but can flexibly adapt to the continuous changes in factors such as material conveying flow, airflow pressure, and particle characteristics in the production process. This means that when production conditions change, the standard for judging whether the acoustic mode is consistent will also be adjusted accordingly, making the judgment more in line with actual operating conditions and avoiding false positives due to changes in operating conditions, thereby enhancing the robustness of the system. Finally, by analyzing whether the mode similarity is greater than the dynamically acquired preset mode similarity threshold, and determining whether the change mode of the first acoustic propagation path is consistent with the preset acoustic propagation path change mode. The accuracy of this final judgment link is guaranteed by the dynamic threshold adjustment. By using a threshold that matches the current operating conditions, the system can more accurately distinguish between changes in acoustic characteristics caused by normal operation and abnormal changes caused by potential faults, thereby effectively reducing the false positive rate and the false negative rate. This dynamic threshold adjustment mechanism makes the entire acoustic propagation path change mode consistency judgment link more intelligent and accurate, thereby improving the overall performance of the entire pharmaceutical equipment fault detection method, allowing it to maintain high levels of detection accuracy and reliability in complex and variable production environments.

[0036] In one specific embodiment, the preset mode similarity threshold can be obtained by pre-establishing a multi-dimensional lookup table or training a machine learning model. The lookup table or model can be constructed based on a large amount of historical production data, which covers the range of normal acoustic propagation mode similarity under different conveying flow fluctuations, airflow pressure fluctuations and particle characteristics. For example, when the conveying flow fluctuation is large, the airflow pressure is low, and the particle size is small, the similarity under the normal mode can be relatively low, and the system can obtain a lower threshold from the lookup table; on the contrary, when the working condition is stable and the particle size is large, the similarity under the normal mode can be higher, and the system can obtain a higher threshold. In actual operation, the system can query the lookup table or input into the trained model according to the real-time conveying flow fluctuation information, airflow pressure fluctuation information and particle characteristics, so as to dynamically output the most suitable preset mode similarity threshold under the current working condition. Finally, when analyzing whether the mode similarity is greater than the preset mode similarity threshold obtained dynamically, the system can simply perform a numerical comparison operation. If the calculated mode similarity value is greater than the obtained dynamic threshold, the system can determine that the change mode of the first acoustic propagation path is consistent with the preset acoustic propagation path change mode, indicating that the current acoustic response conforms to the expected change under normal operation. On the contrary, if the mode similarity is not greater than the threshold, the system can determine that the modes are inconsistent, which may indicate that there is a potential anomaly or fault, which needs further analysis and diagnosis.

[0037] In some preferred embodiments, step S33 comprises: S331, obtaining a preliminary similarity according to the enhanced acoustic propagation path change mode and the preset acoustic propagation path change mode; S332, obtaining environmental parameter information, the environmental parameter information comprising environmental temperature and environmental vibration data; S333, adjusting the preliminary similarity according to the environmental parameter information to obtain the mode similarity.

[0038] In order to overcome the interference of environmental factors on the judgment of the acoustic propagation path mode, a dynamic adjustment mechanism of environmental parameters is introduced in the process of obtaining the mode similarity. First, by obtaining the enhanced acoustic propagation path change mode and the preset acoustic propagation path change mode, the system can preliminarily quantify the similarity between the two, obtaining a preliminary similarity. The preliminary similarity refers to the quantified similarity obtained by directly comparing the current enhanced acoustic propagation path change mode with the preset acoustic propagation path change mode without considering the influence of environmental factors. This preliminary similarity reflects the closeness of the current acoustic characteristics to the normal mode under ideal conditions. However, considering the significant influence of environmental temperature and environmental vibration and other factors on acoustic signal propagation and sensor acquisition in actual production environment, the system further obtains these environmental parameter information. The environmental parameter information refers to the key external environmental data that affects acoustic signal propagation and sensor acquisition, which can be obtained by a special environmental sensor or a monitoring module of existing equipment. Changes in environmental temperature will directly affect the speed of sound and the acoustic characteristics of materials, while environmental vibration may introduce additional noise or change the structural response of the pipeline, which may cause the preliminary similarity to deviate. Therefore, after obtaining the preliminary similarity, the system uses the obtained environmental parameter information to correct the preliminary similarity, for example, compensating for the difference in signal arrival time caused by the change in temperature, or weighting or correcting the similarity calculation caused by the increase in signal noise according to the vibration data. Through this dynamic adjustment based on environmental parameters, the interference of environmental factors on the evaluation of acoustic propagation path characteristics can be effectively eliminated or weakened, making the finally obtained mode similarity more accurately reflect the real changes of the acoustic characteristics in the pipeline. The mode similarity refers to the quantified value that more accurately reflects the real similarity between the current enhanced acoustic propagation path change mode and the preset acoustic propagation path change mode after environmental parameter correction. This method, combined with the previous steps of multi-scale decomposition of the first acoustic propagation path change mode and suppression of transient noise and non-blocking interference, ensures the accuracy and reliability of the mode comparison, so as to more accurately judge whether the current acoustic propagation path change is indeed caused by normal operation, avoid misjudgment caused by environmental fluctuations, and significantly improve the robustness of fault detection.

[0039] In a specific embodiment, the preliminary similarity can be obtained by a feature vector based similarity calculation method, for example, the enhanced acoustic propagation path variation pattern and the preset acoustic propagation path variation pattern are represented as multi-dimensional feature vectors respectively, and then the cosine similarity or Euclidean distance between the two vectors is calculated. The acquisition of environmental parameter information can be achieved by deploying temperature sensors and vibration sensors near the pharmaceutical equipment pipeline. Subsequently, the preliminary similarity is adjusted according to the environmental parameter information to obtain the final pattern similarity, for example, a similarity correction model based on environmental parameters (a polynomial regression model) is established in advance, the preliminary similarity is the dependent variable of the model, and the environmental temperature and environmental vibration data are the independent variables of the model. The model can be trained to obtain a correction function according to historical data, for example, when the environmental temperature rises, a temperature compensation factor is applied to fine-tune the preliminary similarity; when the environmental vibration intensity exceeds a certain threshold, a vibration correction factor is applied to reduce or adjust the weight of the preliminary similarity, so as to reduce the influence of noise on the judgment. In this way, the system can dynamically adapt to environmental changes to ensure the accuracy of pattern similarity calculation.

[0040] In some preferred embodiments, step S5 comprises: S51, obtaining the acoustic characteristic pattern corresponding to the preset non-blocking local structure variation according to the preset equipment geometric topology data and real-time operation parameters; S52, performing preliminary pattern analysis on the feature variation pattern of different pickup sensor pairs according to the path feature difference and the preset equipment geometric topology to obtain an initial abnormal pattern; S53, performing multi-dimensional comparison between the initial abnormal pattern and the acoustic characteristic pattern corresponding to the non-blocking local structure variation, and analyzing whether the initial abnormal pattern is derived from pipeline blockage according to the comparison result; S54, when the initial abnormal pattern is derived from pipeline blockage, identifying the acoustic characteristic abnormal area and locating the fault point according to the initial abnormal pattern.

[0041] The acoustic characteristic pattern corresponding to the preset non-blocking local structure variation refers to a set of acoustic signal features pre-established by experiment, simulation or historical data analysis when the equipment is normally running but there is a non-blocking local structure variation (such as pipeline slight deformation, inner wall wear or temporary material attachment), which can be represented by an acoustic fingerprint library, a parameter model or a machine learning model. The initial abnormal pattern refers to a set of acoustic abnormal signal features obtained by preliminary pattern analysis without classification and verification, which can be represented by a feature vector, a time-frequency graph or frequency spectrum data. The multi-dimensional comparison refers to comparing and analyzing two or more acoustic characteristic patterns in multiple dimensions to evaluate their similarity or difference, which can be achieved by correlation analysis, distance measurement (such as Euclidean distance, Mahalanobis distance) or pattern recognition algorithm (such as support vector machine, neural network).

[0042] The operational logic of this solution begins by acquiring pre-defined acoustic signature patterns corresponding to non-blocking local structural changes based on pre-set equipment geometry and real-time operating parameters. This establishes a baseline for identifying acoustic changes not caused by actual blockage. This step enables the system to pre-identify the acoustic "fingerprints" that may arise from subtle structural changes during normal pipeline operation, providing a critical reference for subsequent anomaly detection. Based on this, a preliminary pattern analysis of the signature change patterns of different pickup sensor pairs is performed based on path feature differences and the pre-set equipment geometry to determine an initial anomaly pattern. This preliminary analysis leverages the path feature differences obtained in the previous step—i.e., abnormal signals compared to a healthy baseline—and combines them with the physical layout of the equipment to preliminarily identify all detected acoustic anomalies, forming an unclassified anomaly set. Subsequently, a multi-dimensional comparison is performed between the initial anomaly pattern and the acoustic signature patterns corresponding to non-blocking local structural changes. The comparison results are used to determine whether the initial anomaly pattern is due to pipeline blockage. This is the key to this solution, as it accurately compares the initially identified anomaly with the pre-established non-blocking change pattern. It is precisely because of this comparison mechanism that the system can determine whether the currently detected anomaly is consistent with the known acoustic change pattern caused by normal operation. If the comparison results show that the initial abnormal pattern is consistent with or highly similar to the non-blocking change pattern, it is determined that the anomaly is not caused by pipeline blockage; on the contrary, if there is a significant difference, it indicates that the anomaly is more likely to be caused by actual pipeline blockage. This judgment mechanism solves the problem of false alarms caused by the difficulty in distinguishing between normal operation changes and actual blockages in the background technology. Ultimately, when the initial abnormal pattern is caused by pipeline blockage, the abnormal area of ​​acoustic characteristics is identified and the fault point is located based on the initial abnormal pattern. Only when the system clearly determines that the anomaly is indeed caused by pipeline blockage will the abnormal pattern be further used to identify the specific abnormal area of ​​acoustic characteristics and accurately locate the fault point. This conditional execution ensures that the subsequent fault location is targeted at the real blockage event, avoiding false alarms and mislocations of non-blocking changes. Through the organic combination of the above steps, this solution can distinguish between acoustic anomalies caused by real blockage and non-blocking local structural changes, thereby improving the accuracy and reliability of fault detection.

[0043] In a specific embodiment, the pharmaceutical equipment fault detection method can be implemented as follows: In the process of obtaining acoustic characteristic patterns corresponding to preset non-blocking local structural changes, high-accuracy displacement sensors and temperature sensors are installed on the pipeline system in advance to monitor the subtle deformation and local temperature changes of the pipeline. Meanwhile, a layer of peelable tracer material is coated on the inner wall of the pipeline, and the degree of material wear is periodically detected to evaluate the wear of the inner wall. For temporary material adhesion, micro-vision sensors or capacitive sensors on the inner wall of the pipeline can be used for detection. When the equipment is running normally but there are these non-blocking changes, acoustic signals are injected by acoustic exciter, and acoustic response signals are collected by pickup sensors. These collected acoustic data, combined with displacement, temperature, wear and adhesion data, can be input into a deep learning model (such as a recurrent neural network or a convolutional neural network) for training, so as to learn and generate an acoustic characteristic pattern library corresponding to different non-blocking local structural changes. For example, when a segment of the pipeline undergoes a subtle deformation of 0.1 millimeter, the amplitude attenuation and phase shift pattern of its acoustic response signal in a certain frequency range can be recorded and used as a preset pattern. In the preliminary pattern analysis to obtain the initial abnormal pattern, the path feature difference data obtained from the previous step S4, such as the acoustic signal arrival time difference between different pickup sensor pairs, the signal waveform distortion degree and the amplitude attenuation amount, can be input into a rule-based expert system or a simple threshold judgment model. For example, if the signal arrival time difference of a certain sensor pair exceeds the preset normal fluctuation range, or its signal attenuation amount increases significantly, the pipeline region corresponding to this sensor pair will be marked as a preliminary abnormal region, and its acoustic characteristic data (such as time-frequency diagram or spectrum) will be output as the initial abnormal pattern. In the multi-dimensional comparison of the initial abnormal pattern with the acoustic characteristic pattern corresponding to the non-blocking local structural change, a feature vector matching algorithm can be used. For example, the acoustic features of the initial abnormal pattern (such as spectral energy distribution, resonance peak frequency, Q value, etc.) are extracted as high-dimensional feature vectors, and the acoustic characteristic pattern corresponding to the preset non-blocking local structural change is also represented as a feature vector. Then, the cosine similarity or Euclidean distance between the two feature vectors can be calculated. If the calculated similarity is higher than a certain preset threshold (for example, 0.9), or the distance is less than a certain preset threshold, it is determined that the initial abnormal pattern is caused by non-blocking local structural changes; otherwise, if the similarity is lower or the distance is larger, it is determined that it is caused by pipeline blockage. In addition, support vector machine (SVM) or random forest classifier can also be used, taking the initial abnormal pattern as input, and directly outputting the judgment result of whether it is caused by pipeline blockage through the trained model. When the initial abnormal pattern is caused by pipeline blockage, the acoustic characteristic abnormal region is identified and the fault point is located according to the initial abnormal pattern.Once it is confirmed that the anomaly is caused by a blockage, the system can use the detailed acoustic feature data in the initial anomaly pattern, combined with preset device geometric topology data, to accurately locate the blockage point through acoustic tomography algorithms or algorithms based on the principle of triangular positioning. For example, by analyzing the signal arrival time difference and signal intensity attenuation of multiple sensor pairs, the specific coordinates of the sound source (i.e., the blockage point) in the pipeline network can be deduced. In addition, the three-dimensional model of the pipeline can be used to visualize the positioning results, directly displaying the specific pipeline segment and location where the blockage occurs on the operation interface, for example, as "Main conveying pipeline A segment, 15.3 meters away from the inlet".

[0044] The present scheme can distinguish between acoustic anomalies caused by actual pipeline blockage and acoustic feature patterns caused by non-blocking local structural changes such as pipeline deformation, internal wall wear, or temporary material attachment, by introducing a judgment mechanism for the source of the anomaly pattern. This significantly reduces the false positive rate and avoids misjudging acoustic changes in normal operation as faults, thereby improving the accuracy and reliability of fault detection. By accurately identifying real blockage events, the present scheme reduces unnecessary on-site investigation workload and improves the operation efficiency and maintenance convenience of the pharmaceutical production line.

[0045] In some preferred embodiments, step S54 comprises: S541, obtaining material property data and structural inhomogeneity data of the pipeline network of the granular agent pharmaceutical production device; S542, determining the influence characteristics of the self-structural inhomogeneity of the pipeline network on acoustic signal propagation according to the material property data and the structural inhomogeneity data; S543, compensating the initial anomaly pattern according to the influence characteristics to obtain a corrected anomaly pattern; S544, identifying an acoustic property abnormal area and locating a fault point according to the corrected anomaly pattern.

[0046] The material property data refers to information describing the physical attributes of the materials constituting the pipe network, which can be characterized by parameters such as material type, density, elastic modulus, Poisson's ratio, and sound velocity. The structural heterogeneity data refers to information describing the irregularities of the geometric structure and physical state of the pipe network, which can be characterized by parameters such as pipe wall thickness variation, inner surface roughness, elbow curvature, weld location, valve type, and branch point geometry. The influence feature refers to the specific change pattern of the acoustic signal propagation law caused by the structural heterogeneity of the pipe network itself, which can be represented by parameters such as sound wave reflection coefficient, scattering coefficient, attenuation coefficient, frequency response curve, or energy loss model of a specific frequency band. The compensation processing refers to the operation of eliminating or weakening the interference caused by the structural heterogeneity of the pipe itself contained in the initial abnormal pattern, which can be implemented by techniques such as signal filtering, background noise subtraction, model prediction residual analysis, or pattern separation based on machine learning. The corrected abnormal pattern refers to the pattern after compensation processing, which can more accurately reflect the change of acoustic characteristics caused by pipe blockage, which can be represented by parameters such as residual acoustic feature difference, pure abnormal signal spectrum, or calibrated propagation path feature deviation.

[0047] Based on the above understanding of the key features, the overall operation principle of the method is as follows: First, by obtaining the material property data and structural heterogeneity data of the pipeline network of the granular drug production equipment, a foundation is laid for subsequent accurate analysis. These data comprehensively describe the inherent physical properties and geometric structural irregularities of the pipeline itself, which have an inherent influence on the propagation of acoustic signals, but are not caused by blockage. Based on these detailed material property data and structural heterogeneity data, the system can determine the influence characteristics of the structural heterogeneity of the pipeline network itself on the propagation of acoustic signals. This process is achieved by establishing an accurate model of the inherent structure of the pipeline reflecting, scattering, attenuating or distorting sound waves, thereby forming a characterization of structural noise or background influence. It is precisely because of the establishment of this influence characteristic model that the system can effectively distinguish the acoustic changes caused by the pipeline structure itself from the acoustic anomalies caused by actual blockage. On this basis, the initial anomaly pattern is compensated according to the determined influence characteristics to obtain a corrected anomaly pattern. In the previous scheme, the initial anomaly pattern is obtained based on the preliminary analysis of the path feature differences, which may contain interference caused by the structural heterogeneity of the pipeline itself. Through the compensation processing of the present scheme, the influence of the inherent structure of the pipeline on the propagation of acoustic signals is effectively removed or significantly weakened from the initial anomaly pattern, thereby obtaining a more pure and accurate corrected anomaly pattern. This corrected anomaly pattern can more truly reflect the changes in acoustic characteristics caused by pipeline blockage, thereby effectively reducing the interference of non-blockage structural factors on fault judgment. Finally, the acoustic characteristic abnormal area is identified and the fault point is located according to the corrected anomaly pattern. Since the corrected anomaly pattern has ruled out the influence of the structural heterogeneity of the pipeline itself, it can more accurately indicate the real acoustic abnormal area, thereby achieving accurate positioning of the fault point. The present scheme further refines the processing based on the identification of the initial anomaly pattern, removes the interference caused by the structural heterogeneity of the pipeline itself, so that the subsequent fault positioning is no longer limited by these non-blockage factors, thereby significantly improving the reliability and efficiency of fault diagnosis.

[0048] In one specific embodiment, the method can be implemented as follows: first, material property data and structural heterogeneity data of the granular pharmaceutical production equipment pipeline network are obtained from a structured database. Then, based on the material property data and structural heterogeneity data, the influence of the structural heterogeneity of the pipeline network itself on acoustic signal propagation is determined. For example, finite element analysis (FEA) software can be used to simulate the propagation of sound waves in the pipeline based on the geometric model and material properties of the pipeline, calculate the reflection, transmission and attenuation characteristics of sound waves at structures such as bends, welds, reducers, etc. at different frequencies, and generate a feature library of acoustic propagation influences. This feature library can include the influence of specific structures on the arrival time, amplitude, phase and spectral shape of the acoustic signal. Then, the initial anomaly pattern is compensated based on the determined influence features to obtain a corrected anomaly pattern. For example, when the system detects an initial anomaly pattern, the influence features corresponding to the structure of the current pipeline segment can be retrieved from the pre-established influence feature library. Subsequently, signal processing algorithms such as adaptive filtering or model-based residual analysis can be used to subtract or weightedly suppress these influence features from the initial anomaly pattern. For example, if a bend causes attenuation of sound waves at a specific frequency, the attenuation effect caused by the bend can be compensated in the opposite direction in the initial anomaly pattern, resulting in a corrected anomaly pattern that excludes the influence of the bend. Finally, the acoustic property abnormal area is identified and the fault point is located based on the corrected anomaly pattern. For example, the processing system can perform further pattern recognition and spatial mapping on the corrected anomaly pattern. By comparing the corrected anomaly pattern with the baseline pattern in the non-clogging state and combining the geometric topology of the pipeline, the area where the acoustic properties change significantly can be accurately identified and mapped to the specific physical location in the pipeline network, thereby locating the clogging fault point. For example, if the corrected anomaly pattern shows that the sound speed of a straight pipe segment is significantly reduced and the amplitude attenuation anomaly is increased, the fault point can be accurately located to the straight pipe segment.

[0049] In some preferred embodiments, the pre-construction process of the pipeline acoustic propagation path feature spectrum baseline database includes: obtaining preset equipment geometric topology data of the granular pharmaceutical production equipment and installation position information of the acoustic exciter and the plurality of pickup sensors; under a plurality of production running states of the pipeline of the granular pharmaceutical production equipment without clogging, injecting acoustic signals into the pipeline through the acoustic exciter, synchronously collecting acoustic response signals of the pipeline by the plurality of pickup sensors, and recording corresponding production running parameters; processing the acoustic response signals to extract multi-dimensional acoustic propagation path features, the multi-dimensional acoustic propagation path features including arrival time differences of acoustic signals between different sensor pairs, signal waveform distortion degrees, amplitude attenuation amounts of specific frequencies, phase shift amounts, bandwidths of resonance peaks, and damping factors; The multi-dimensional acoustic propagation path characteristics are associated with the corresponding pipeline sections, sensor pairs, and recorded production operation parameters to construct a baseline database of pipeline acoustic propagation path characteristic spectra.

[0050] Multidimensional acoustic propagation path features are a collection of independent or correlated parameters used to comprehensively describe the propagation characteristics of sound waves within a pipeline. These features can be extracted by analyzing the acquired acoustic response signals in the time domain, frequency domain, or time-frequency domain. For example, these features can be obtained by calculating the signal's cross-correlation function, Fourier transform, wavelet transform, or Hilbert transform. The combination of these features can more precisely reflect the state of the medium and structural changes within the pipeline, thereby improving the ability to identify anomalies.

[0051] Second, as Figure 2 As shown, the present application also provides a pharmaceutical equipment fault detection system, which includes: Information acquisition module 1 is used to obtain real-time operating parameters of the granule production equipment during the granule production process, and inject acoustic signals through an acoustic actuator and collect acoustic response signals at different positions of the granule production equipment through multiple sound pickup sensors; the real-time operating parameters include material conveying flow rate, air flow pressure, and the activation / deactivation status of each branch pipeline; Path feature acquisition module 2, used to extract the first acoustic propagation path feature according to all acoustic response signals; Analysis module 3, used to obtain a preset acoustic propagation path change pattern according to real-time operating parameters, and analyze whether the change pattern of the first acoustic propagation path is consistent with the preset acoustic propagation path change pattern; a difference acquisition module 4 for comparing the first acoustic propagation path characteristic with a second acoustic propagation path characteristic corresponding to the real-time operating parameters in a pre-established pipeline acoustic propagation path characteristic spectrum baseline database to obtain a path characteristic difference when the change pattern of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change pattern; The fault analysis module 5 is used to analyze the characteristic change patterns of different sound pickup sensor pairs based on the path characteristic differences and the preset equipment geometric topology data to identify the abnormal area of ​​acoustic characteristics and locate the fault point.

[0052] A pharmaceutical equipment fault detection system provided in the present application includes an information acquisition module 1, a path feature acquisition module 2, an analysis module 3, a difference acquisition module 4 and a fault analysis module 5. The pharmaceutical equipment fault detection system provided in this embodiment is used to execute the steps in the pharmaceutical equipment fault detection method provided in the first aspect above. The principle of the pharmaceutical equipment fault detection system provided in this embodiment is the same as the principle of the pharmaceutical equipment fault detection method provided in the first aspect above, and will not be discussed in detail here.

[0053] From the above, the pharmaceutical equipment fault detection method and system provided by the application can realize the identification and positioning of the pharmaceutical equipment fault by comprehensively utilizing the real-time operation parameters, the multi-sensor acoustic data and the equipment topology structure information, and therefore the application can effectively solve the problem that the existing pharmaceutical equipment fault detection system cannot locate the fault point in a complex pipeline network and overcome the challenge of being difficult to distinguish the changes caused by normal operation and actual faults, thereby improving the accuracy and efficiency of fault diagnosis and reducing the time and cost of manual troubleshooting.

[0054] In the embodiments provided in the application, it should be understood that the disclosed devices and methods can be implemented in other manners. The above-described device embodiments are merely illustrative. For example, the division of the units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another robot, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between the units can be indirect coupling or communication connection through some communication interfaces, and can be electrical, mechanical or other forms.

[0055] In addition, each functional module in each embodiment of the application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0056] In this document, the terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations.

[0057] The above is only an embodiment of the application and is not used to limit the protection scope of the application. For those skilled in the art, the application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A pharmaceutical equipment fault detection method, characterized in that: The pharmaceutical equipment fault detection method includes the following steps: S1. During the granule drug production process, obtain real-time operating parameters of the granule drug production equipment, inject acoustic signals through an acoustic actuator, and collect acoustic response signals at different locations of the granule drug production equipment through multiple sound pickup sensors; the real-time operating parameters include material conveying flow rate, air flow pressure, and the activation / deactivation status of each branch pipeline; S2. extracting a first acoustic propagation path feature according to all acoustic response signals; S3. Obtaining a preset acoustic propagation path change pattern according to the real-time operating parameters, and analyzing whether the change pattern of the first acoustic propagation path is consistent with the preset acoustic propagation path change pattern; S4. When the change pattern of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change pattern, comparing the first acoustic propagation path characteristics with the second acoustic propagation path characteristics corresponding to the real-time operating parameters in a pre-established pipeline acoustic propagation path characteristic spectrum baseline database to obtain a path characteristic difference; S5. Analyze the characteristic change patterns of different sound pickup sensor pairs based on the path characteristic differences and the preset device geometric topology data to identify the abnormal area of ​​acoustic characteristics and locate the fault point.

2. The pharmaceutical equipment fault detection method according to claim 1, characterized in that: The first acoustic propagation path characteristic and the second acoustic propagation path characteristic both include the arrival time difference of the acoustic signals between different pickup sensor pairs, the degree of signal waveform distortion, the amplitude attenuation, the phase shift, and the bandwidth and damping factor of the resonance peak.

3. The pharmaceutical equipment fault detection method according to claim 1, characterized in that: Step S3 includes: S31. Performing multi-scale decomposition on the variation pattern of the first acoustic propagation path, and combining it with real-time operating parameters, identifying and suppressing transient noise and non-blocking interference caused by particle collisions and airflow pulsations, to obtain an enhanced acoustic propagation path variation pattern. S32. Obtaining a preset acoustic propagation path change pattern according to real-time operating parameters; S33. Obtaining pattern similarity according to the enhanced acoustic propagation path change pattern and the preset acoustic propagation path change pattern; S34. Analyze whether the pattern similarity is greater than the preset pattern similarity threshold. If so, determine that the change pattern of the first acoustic propagation path is consistent with the preset acoustic propagation path change pattern. If not, determine that the change pattern of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change pattern.

4. The pharmaceutical equipment fault detection method according to claim 3, characterized in that: Step S31 includes: S311. Perform multi-scale decomposition on the change pattern of the first acoustic propagation path to obtain acoustic components of different time scales and frequency ranges; S312. Obtaining characteristics of particles currently produced by the granule drug production equipment; the particle characteristics include particle type and particle size distribution; S313, obtaining particle collision noise characteristics and airflow pulsation noise characteristics according to particle characteristics and real-time operating parameters; S313, identifying, based on the particle collision noise characteristics and the airflow pulsation noise characteristics, portions of the acoustic components corresponding to transient noise and non-blocking interference caused by particle collision and airflow pulsation; S314. Suppress the portion of the acoustic component corresponding to the transient noise and non-blocking interference caused by particle collision and airflow pulsation, and reconstruct and generate an enhanced acoustic propagation path change pattern based on the suppressed acoustic component.

5. The pharmaceutical equipment fault detection method according to claim 3, characterized in that: Step S34 includes: S341, obtaining the delivery flow rate fluctuation information and the air flow pressure fluctuation information according to the real-time operating parameters, and obtaining the characteristics of the particles currently produced by the granule drug production equipment; S342, obtaining a preset pattern similarity threshold based on the delivery flow fluctuation information, the airflow pressure fluctuation information, and the particle characteristics; S343. Analyze whether the pattern similarity is greater than the preset pattern similarity threshold. If so, determine that the change pattern of the first acoustic propagation path is consistent with the preset acoustic propagation path change pattern. If not, determine that the change pattern of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change pattern.

6. The pharmaceutical equipment fault detection method according to claim 3, characterized in that: Step S33 includes: S331. Obtaining a preliminary similarity based on the enhanced acoustic propagation path change pattern and the preset acoustic propagation path change pattern; S332. Acquire environmental parameter information, where the environmental parameter information includes ambient temperature and ambient vibration data; S333. Adjust the preliminary similarity according to the environmental parameter information to obtain pattern similarity.

7. The pharmaceutical equipment fault detection method according to claim 1, characterized in that: Step S5 includes: S51. Obtaining a preset acoustic characteristic pattern corresponding to a non-blocking local structural change based on preset equipment geometric topology data and real-time operating parameters; S52, performing preliminary pattern analysis on characteristic change patterns of different sound pickup sensor pairs based on path characteristic differences and a preset device geometric topology to obtain an initial abnormal pattern; S53, performing a multi-dimensional comparison between the initial abnormal pattern and the acoustic characteristic pattern corresponding to the non-blocking local structural change, and analyzing whether the initial abnormal pattern is caused by the pipe blockage based on the comparison results; S54. When the initial abnormal pattern is caused by pipeline blockage, identify the abnormal area of ​​acoustic characteristics and locate the fault point according to the initial abnormal pattern.

8. The pharmaceutical equipment fault detection method according to claim 7, characterized in that: Step S54 includes: S541. Obtain material property data and structural heterogeneity data of the pipeline network of the granule drug production equipment; S542. Determine, based on the material property data and the structural nonuniformity data, the characteristics of the influence of the structural nonuniformity of the pipeline network on the propagation of the acoustic signal; S543, performing compensation processing on the initial abnormal pattern according to the influencing characteristics to obtain a corrected abnormal pattern; S544: Identify the abnormal area of ​​acoustic characteristics and locate the fault point according to the corrected abnormal pattern.

9. The pharmaceutical equipment fault detection method according to claim 1, characterized in that: The pre-construction process of the pipeline acoustic propagation path characteristic spectrum baseline database includes: Obtaining preset equipment geometric topology data of a granule drug production equipment and installation position information of an acoustic actuator and a plurality of pickup sensors; Under various production operation conditions of the granule drug production equipment with no pipeline blockage, an acoustic signal is injected into the pipeline through an acoustic actuator. Multiple pickup sensors synchronously collect the pipeline's acoustic response signals and record the corresponding production operation parameters. Process the acoustic response signal to extract multidimensional acoustic propagation path characteristics, which include the arrival time difference of acoustic signals between different sensor pairs, the degree of signal waveform distortion, the amplitude attenuation at a specific frequency, the phase offset, the bandwidth of the resonance peak, and the damping factor; The multi-dimensional acoustic propagation path characteristics are associated with the corresponding pipeline sections, sensor pairs, and recorded production operation parameters to construct a baseline database of pipeline acoustic propagation path characteristic spectra.

10. A pharmaceutical equipment fault detection system, characterized in that: Pharmaceutical equipment fault detection system includes: An information acquisition module is used to obtain real-time operating parameters of the granule production equipment during the granule production process, injecting acoustic signals through an acoustic actuator and collecting acoustic response signals at different locations of the granule production equipment through multiple sound pickup sensors; the real-time operating parameters include material conveying flow rate, airflow pressure, and the activation / deactivation status of each branch pipeline; A path feature acquisition module, configured to extract a first acoustic propagation path feature based on all acoustic response signals; an analysis module, configured to obtain a preset acoustic propagation path change pattern based on real-time operating parameters, and analyze whether the change pattern of the first acoustic propagation path is consistent with the preset acoustic propagation path change pattern; a difference acquisition module, configured to compare the first acoustic propagation path characteristics with the second acoustic propagation path characteristics corresponding to the real-time operating parameters in a pre-built pipeline acoustic propagation path characteristic spectrum baseline database to obtain a path characteristic difference when the change pattern of the first acoustic propagation path is inconsistent with the preset acoustic propagation path change pattern; The fault analysis module is used to analyze the characteristic change patterns of different sound pickup sensor pairs based on path characteristic differences and preset equipment geometric topology data to identify abnormal areas of acoustic characteristics and locate fault points.

Images