A method and system for detecting the state of a device for PCR
By constructing an adaptive clustering model and a Gaussian mixture model, the problem of monitoring blind spots in the liquid flow control of PCR equipment was solved, enabling timely and accurate detection of local anomalies, improving the detection accuracy and response speed of PCR equipment, and reducing the waste of samples and reagents.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NINGBO INARRAY BIOMEDICAL SYST CO LTD
- Filing Date
- 2026-02-24
- Publication Date
- 2026-05-29
AI Technical Summary
Existing PCR equipment has monitoring blind spots in liquid flow control, making it difficult to detect local abnormal operating conditions such as blockages and leaks in a timely and accurate manner, resulting in sample loss and reagent waste. This is especially true in complex nucleic acid testing processes where the detection accuracy is insufficient and the response delay is severe.
By acquiring the historical operating parameters of the drive motors in each stage, an adaptive clustering model is constructed. A Gaussian mixture model is used to independently detect the abnormal state of each stage. By combining the abnormal judgment threshold and the state judgment index, accurate state detection of the PCR equipment is achieved.
It significantly improves the accuracy and timeliness of anomaly detection, reduces sample and reagent waste, and ensures the reliability of PCR test results, making it particularly suitable for PCR equipment used for rapid on-site testing.
Smart Images
Figure CN121705871B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electronic digital data processing technology, and in particular to a method and system for detecting the status of PCR equipment. Background Technology
[0002] PCR (Polymerase Chain Reaction) technology is one of the core technologies in modern molecular biology detection, playing an irreplaceable role, especially in food safety (such as identification of meat-derived components) and disease diagnosis. To meet the needs of rapid, on-site testing, integrated PCR devices have been widely used. Their core component is the microfluidic chip, which integrates multiple functional units such as sample processing, nucleic acid extraction, purification, amplification, and detection onto a tiny platform, achieving automation and miniaturization of the entire detection process from sample processing to result analysis. In typical microfluidic chip-based PCR devices, the microfluidic chip serves as the core carrier, integrating modules such as a sample dispensing port, purification reagent storage chamber, nucleic acid extraction and purification chamber, waste liquid chamber, and real-time fluorescence PCR amplification chamber. Precise liquid actuation and control are crucial to ensuring the accuracy of test results.
[0003] Liquid flow control technology in PCR equipment typically employs a single motor to drive a rotating pump ring on a microfluidic chip. The rotation of this ring sequentially activates multiple micropumps integrated into the microfluidic chip, propelling liquid through various functional compartments along a pre-defined path. Due to physical space and cost limitations, it's difficult to deploy independent flow or pressure sensors at each micropump location. Therefore, while this centralized drive scheme simplifies the hardware structure, it introduces significant monitoring blind spots, making it impossible to directly and independently acquire real-time status data for each micropump. During PCR equipment operation, these monitoring blind spots mean that existing equipment operation status detection methods struggle to provide timely and accurate feedback on abnormal operating conditions such as blockages, leaks, or sudden flow changes. This can easily lead to sample loss, reagent waste, and even distorted PCR test results.
[0004] Specifically, current monitoring of PCR equipment operation relies primarily on simple threshold judgments of overall drive motor operating parameters (such as total current and average speed). However, this method for detecting abnormal PCR equipment operation has significant shortcomings: First, the overall drive motor operating parameters cannot reflect anomalies in individual micropumps or specific flow channels, such as local blockages, micro-leakage, or air bubble intrusion. Abnormal signals are easily overwhelmed by the overall data, leading to insufficient accuracy in operation status detection and a high risk of missed detections. Furthermore, anomalies can only be detected when they accumulate to a certain level and affect overall motor performance, resulting in significant response delays that may lead to sample or reagent waste or even invalid PCR test results. This is particularly problematic for complex multi-step, multi-reagent nucleic acid extraction and purification processes, such as meat nucleic acid testing. PCR equipment operation involves multiple steps, including sample loading, washing, elution, and amplification. The fluid characteristics and control precision requirements of each step differ, making it difficult to adapt a uniform monitoring threshold to the characteristics of different steps, easily leading to false alarms or missed detections. Summary of the Invention
[0005] To achieve timely detection and accurate assessment of abnormal operating conditions of PCR equipment in the complex meat nucleic acid testing process, and to overcome the shortcomings of existing methods such as insufficient detection accuracy and severe response delay, this invention provides a method and system for PCR equipment status detection, the technical solution of which is as follows:
[0006] In a first aspect, the present invention provides a method for detecting the equipment status of PCR, comprising the following steps: acquiring historical normal operating parameters of the drive motors at each stage, and dividing the data into sets of status data segments corresponding to each stage; constructing historical equipment status vector sets for each stage based on the sets of status data segments for each stage; calculating anomaly judgment thresholds for each stage based on the historical equipment status vector sets for each stage; constructing an adaptive clustering model for each stage based on a Gaussian mixture model; acquiring real-time data of the target PCR equipment, and calculating the status judgment index for the current stage based on the corresponding adaptive clustering model after the current stage is completed; and detecting and judging the equipment status of the current stage based on the status judgment index and the corresponding anomaly judgment threshold.
[0007] Among them, based on several PCR devices of the same model as the target PCR device, in multiple identical historical normal meat nucleic acid testing processes, the historical normal operation parameters of the rotary pump ring drive motor at each stage were collected synchronously at a fixed sampling frequency, including torque data, speed data, and current data, to obtain multiple three-dimensional historical normal sequences corresponding to each stage.
[0008] Preferably, the process involves obtaining the process labels and runtime of each step from the control system of the PCR equipment, obtaining the process label and runtime corresponding to each three-dimensional historical normal sequence, and calculating the total runtime of each historical normal meat nucleic acid testing process. Each step is then numbered sequentially according to its order, and the three-dimensional historical normal sequences are categorized according to their process labels. The combination of the three-dimensional historical normal sequence and the corresponding process label for each step is used as the set of status data segments corresponding to each step.
[0009] Preferably, the maximum historical normal torque, maximum historical normal speed, and maximum historical normal current are extracted from all historical normal operating parameters. For any three-dimensional historical normal sequence in the state data segment set of a certain link, the ratio between the mean of each data point in the torque historical normal sequence and the maximum historical normal torque is taken as the first flow component; the ratio between the variance of each data point in the speed historical normal sequence and the square of the maximum historical normal speed is taken as the second flow component; the ratio between the mean of each data point in the current historical normal sequence and the maximum historical normal current is taken as the first system component; and the coefficient of variation of the current historical normal sequence is taken as the second system component. The four components corresponding to the same three-dimensional historical normal sequence are merged into a historical equipment state vector, and the historical equipment state vector set of each link is obtained in the same way.
[0010] Preferably, any link is selected as the target link, and the runtime corresponding to each three-dimensional historical normal sequence under the target link is extracted. The average of the ratios between each runtime and the corresponding total runtime is used as the importance weight of the target link. The value of 1 minus the importance weight is used as the first sensitivity coefficient of the target link. Similarly, the first sensitivity coefficient of each link is obtained. The historical equipment state vector set of the target link is extracted, and the four components of each historical equipment state vector are extracted to form four component sequences. The coefficient of variation of each of the four component sequences is calculated, and the average of the four coefficients of variation is used as the anomaly tolerance of the target link. Similarly, the anomaly tolerance of each link is obtained. Then, the anomaly tolerance of each link is normalized, and the value of the normalized anomaly tolerance is used as the second sensitivity coefficient of the corresponding link. The product between the first sensitivity coefficient and the corresponding second sensitivity coefficient of each link is used as the anomaly tolerance index of each link.
[0011] Preferably, based on the detection accuracy requirements of the actual application scenario and human experience, a safety threshold for the anomaly tolerance index is set. When the value of the anomaly tolerance index of a certain link exceeds the safety threshold, the safety threshold is used as the value of the anomaly tolerance index of that link. Based on the life cycle and operating wear and tear of the target PCR equipment, a benchmark anomaly judgment threshold is set. The product between the anomaly tolerance index of the target link and the benchmark anomaly judgment threshold is used as the anomaly tolerance increment of the target link. The sum of the anomaly tolerance increment of the target link and the benchmark anomaly judgment threshold is used as the anomaly judgment threshold of the target link. Similarly, the anomaly judgment thresholds of each link are obtained.
[0012] Preferably, a Gaussian mixture model is used to model the probability distribution of each stage, setting a range for the number of Gaussian distributions, and substituting each value into the Gaussian mixture model in turn. For any stage's Gaussian mixture model, a certain value of the number of Gaussian distributions is substituted, and based on the historical equipment state vector set of that stage, the expectation-maximization algorithm is used to fit the Gaussian mixture model. To avoid local optima, multiple random initializations are performed to obtain multiple initial models corresponding to the number of Gaussian distributions. Based on the existing probability density function, the log-likelihood value of each initial model is calculated, and the initial model with the largest log-likelihood value is selected as the initial clustering model corresponding to the number of Gaussian distributions in the corresponding stage. Similarly, the initial clustering model corresponding to each value of the number of Gaussian distributions in each stage is obtained. The optimal number of Gaussian distributions in each stage is determined using the Bayesian information criterion, thereby obtaining the adaptive clustering model for each stage.
[0013] Preferably, starting from the beginning of the current meat nucleic acid testing process, the operating parameters of the rotary pump ring drive motor in the target PCR device are collected synchronously in real time at the same sampling frequency. After the current step is completed, the current device state vector corresponding to the current step is constructed through the same steps. The current device state vector is input into the adaptive clustering model corresponding to the current step to obtain the log-likelihood value of the current step. The historical maximum log-likelihood value and historical minimum log-likelihood value corresponding to the current step are extracted based on the historical device state vector set. The difference between the historical maximum log-likelihood value and the historical minimum log-likelihood value is used as the normal benchmark. The difference between the historical maximum log-likelihood value and the log-likelihood value of the current step is used as the abnormal deviation feature of the current step. The ratio between the abnormal deviation feature and the normal benchmark is used as the state judgment index of the current step.
[0014] Preferably, the status judgment index of the current stage is compared with the anomaly judgment threshold corresponding to the current stage. When the status judgment index is less than the anomaly judgment threshold, the target PCR device is determined to be operating normally in the current stage. When the status judgment index is greater than or equal to the anomaly judgment threshold, the target PCR device is determined to have an abnormal operating state in the current stage and an anomaly alarm is triggered.
[0015] Secondly, the present invention provides a device status detection system for PCR, for implementing the above-mentioned device status detection method for PCR, comprising: a processor, a memory, a communication interface, an early warning device, and a PCR device control system. The processor stores computer program instructions for implementing the above-mentioned device status detection method for PCR, and the communication interface is communicatively connected to the PCR device control system and the early warning device.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0017] This invention maps the liquid flow state, which is difficult to monitor directly on a microfluidic chip, to easily obtainable driving motor operating parameters. It innovatively constructs a normal state benchmark and adaptive anomaly detection threshold independently for each step in the detection process based on a Gaussian mixture model. This effectively overcomes the monitoring blind spot problem caused by the inability to deploy sensors for each micropump in existing technologies. It achieves a leap from overall, coarse judgment to local, refined perception, significantly improving the accuracy and timeliness of anomaly detection. Simultaneously, the adaptive threshold mechanism enhances the adaptability to the differences in characteristics of each step, enabling accurate early warning when anomalies occur in early stages, thereby reducing sample and reagent waste and ensuring the reliability of PCR test results. It is particularly suitable for handheld and other on-site rapid PCR testing devices. Attached Figure Description
[0018] Figure 1 This is a flowchart illustrating an implementation method for equipment status detection in PCR according to an embodiment of the present invention.
[0019] Figure 2 This is a structural block diagram of a device status detection system for PCR according to an embodiment of the present invention. Detailed Implementation
[0020] The technical features of the present invention will be further described in detail below with reference to the accompanying drawings so that those skilled in the art can understand them.
[0021] A method for detecting the status of equipment used in PCR, the implementation process is as follows: Figure 1 As shown, the specific implementation steps are as follows:
[0022] Step S1: Obtain the historical normal operating parameters of the drive motor under each stage, and divide the corresponding state data segment set for each stage.
[0023] Specifically, based on several PCR devices of the same model as the target PCR device, historical normal operating parameters of the rotary pump ring drive motor at each stage are synchronously collected at a fixed sampling frequency during multiple identical historical normal meat nucleic acid testing processes. These parameters include torque data, speed data, and current data, resulting in multiple three-dimensional historical normal sequences corresponding to each stage. The stage labels and runtime of each stage are obtained from the control system of the PCR device. The stage label and runtime corresponding to each three-dimensional historical normal sequence are obtained, and the total runtime of each historical normal meat nucleic acid testing process is calculated. Each stage is numbered sequentially according to its order, and the three-dimensional historical normal sequences are categorized according to their stage labels. The combination of the three-dimensional historical normal sequence and the corresponding stage label for each stage is used as the set of state data segments corresponding to each stage.
[0024] The sampling frequency can be set to 100Hz, and torque, speed, and current data can be directly obtained from the controller of the drive motor inside the PCR device, i.e., from the control system of the PCR device. The typical process for nucleic acid testing of meat includes: the delivery of the test solution from the sample loading port to the nucleic acid extraction and purification chamber, the nucleic acid extraction and purification process, the delivery of waste liquid to the waste liquid chamber, the delivery of elution buffer to the nucleic acid extraction and purification chamber, and the delivery of the nucleic acid eluted solution to the real-time fluorescence PCR expansion chamber.
[0025] Step S2: Based on the set of status data segments of each stage, construct the historical equipment status vector set of each stage.
[0026] Since the abnormal state characteristics of PCR equipment in the meat nucleic acid testing process are difficult to quantify directly, and there is a corresponding mapping relationship between the liquid flow state in the microfluidic chip and the operating parameters of the drive motor, this step designs a historical equipment state vector to map the historical normal operating parameters of the drive motor into the liquid flow state characteristics under historical normal operating conditions. This quantitatively characterizes the working state of the PCR equipment under historical normal operating conditions, laying a data foundation for the subsequent construction of an adaptive abnormal state detection mechanism.
[0027] Specifically, the maximum historical normal torque, maximum historical normal speed, and maximum historical normal current are extracted from all historical normal operating parameters. For any three-dimensional historical normal sequence in the state data segment set of a certain link, the ratio between the mean of each data point in the torque historical normal sequence and the maximum historical normal torque is taken as the first flow component; the ratio between the variance of each data point in the speed historical normal sequence and the square of the maximum historical normal speed is taken as the second flow component; the ratio between the mean of each data point in the current historical normal sequence and the maximum historical normal current is taken as the first system component; and the coefficient of variation of the current historical normal sequence is taken as the second system component. The four components corresponding to the same three-dimensional historical normal sequence are merged into a historical equipment state vector. Similarly, the historical equipment state vector set of each link is obtained.
[0028] The first flow component, the second flow component, and the first system component are quantized using ratios, which eliminates the influence of dimensions and allows for comparison and calculation between different electrical parameters. The second system component is quantized using the coefficient of variation, which is dimensionless data, thus avoiding dimension issues in subsequent calculations.
[0029] Specifically, the first flow component characterizes the average load intensity borne by the drive motor in the corresponding stage. When the viscosity of the liquid in the flow channel of the microfluidic chip increases or blockage occurs, the overall torque data increases, and the value of the first flow component increases; conversely, when the sealing of the flow channel in the microfluidic chip is damaged or even liquid leakage occurs, the overall torque data decreases, and the value of the first flow component decreases. An excessively large or small first flow component indicates a greater likelihood of blockage or leakage abnormalities in the corresponding stage of the PCR equipment. The second flow component characterizes the stability of the drive motor speed. If abnormal liquid flow occurs, such as bubbles or intermittent blockage, it will lead to increased fluctuations in the drive motor speed. Therefore, the variance of the speed is quantified as the second flow component. The larger the value of the second flow component, the more unstable the liquid flow control state in the corresponding stage, and the greater the likelihood of abnormalities; conversely, the smaller the value of the second flow component, the more stable the liquid flow control state in the corresponding stage.
[0030] Furthermore, the first system component characterizes the average current intensity of the drive motor. When abnormalities such as blockage occur in the flow channels of the microfluidic chip, the load intensity on the drive motor increases. To maintain uniform liquid delivery, the corresponding drive current is adjusted to increase, and the value of the first system component increases significantly. Conversely, when abnormalities such as liquid leakage occur, the load intensity on the drive motor decreases, the corresponding drive current is adjusted to decrease, and the value of the first system component decreases significantly. The second system component characterizes the relative stability of the PCR equipment in the corresponding stage of operation. The smaller the coefficient of variation of the normal current history sequence, the smaller the current fluctuation and the more stable the operation of the corresponding stage. Conversely, the larger the coefficient of variation, the more drastic the current fluctuation and the more likely there is an intermittent abnormality in the corresponding stage. The first and second system components characterize whether abnormalities occur in the corresponding stages from the perspective of the energy consumption of the PCR equipment. Both excessively large and small values of the first system component indicate that abnormalities are more likely to occur in the corresponding stages, while the larger the value of the second system component, the more likely an abnormality has occurred in the system.
[0031] Step S3: Based on the historical device status vector set of each link, calculate the anomaly judgment threshold for each link.
[0032] Because PCR equipment can be affected by factors such as wear and tear and ambient temperature, the judgment criteria built based on historical normal operating parameters may deviate to some extent; therefore, it is necessary to set an adaptive anomaly judgment threshold to achieve high-precision detection of abnormal states of PCR equipment.
[0033] Specifically, any stage is selected as the target stage. The runtime corresponding to each three-dimensional historical normal sequence under the target stage is extracted. The average of the ratios between each runtime and the corresponding total runtime is used as the importance weight of the target stage. The value of 1 minus the importance weight is used as the first sensitivity coefficient of the target stage. Similarly, the first sensitivity coefficient of each stage is obtained. The historical equipment state vector set of the target stage is extracted. The four components of each historical equipment state vector are extracted to form four component sequences. The coefficient of variation of each of the four component sequences is calculated. The average of the four coefficients of variation is used as the anomaly tolerance of the target stage. Similarly, the anomaly tolerance of each stage is obtained. The anomaly tolerance of each stage is then normalized. The value of the normalized anomaly tolerance is used as the second sensitivity coefficient of the corresponding stage. The product of the first sensitivity coefficient and the corresponding second sensitivity coefficient of each stage is used as the anomaly tolerance index of each stage.
[0034] Among them, the anomaly tolerance index of link i is The calculation formula is as follows:
[0035]
[0036] In the formula, This represents the safety threshold for anomaly tolerance indicators. The value can be set to 0.8. This represents the importance weight of stage i, which is the average percentage of the runtime of stage i. Let represent the coefficient of variation of the component sequence composed of the j-th component in each historical device state vector of stage i. To represent the normalization function, the Min-Max normalization function can be used. This represents the function for selecting the minimum value.
[0037] Importance weights characterize the importance of each step in the overall meat nucleic acid testing process. The larger the time taken, the more critical the core function of the corresponding step in the meat nucleic acid testing process. In this case, the higher the importance weight value of the corresponding step and the lower the value of the first sensitivity coefficient, the lower the tolerance of the corresponding step to abnormal conditions. Conversely, the lower the importance weight value and the higher the value of the first sensitivity coefficient, the smaller the impact of abnormalities in the corresponding step on the final nucleic acid test results, and therefore the higher the tolerance of the corresponding step to abnormal conditions. The second sensitivity coefficient characterizes the tolerance of the corresponding step to the intrinsic stability of the PCR equipment. By comprehensively evaluating the coefficient of variation of each component in the historical equipment state vector, the concentration of normal states of the corresponding step is quantified. The smaller the coefficient of variation and the smaller the value of the second sensitivity coefficient, the more consistent the equipment state performance under historical normal operation, the lower the tolerance of the corresponding step to external interference and various influencing factors, and the more stringent abnormal judgment criteria need to be set.
[0038] Furthermore, based on the detection accuracy requirements of actual application scenarios and human experience, a safety threshold for the anomaly tolerance index is set. When the value of the anomaly tolerance index of a certain link exceeds the safety threshold, the safety threshold is used as the value of the anomaly tolerance index of that link. Based on the life cycle and operating wear and tear of the target PCR equipment, a baseline anomaly judgment threshold is set. The product between the anomaly tolerance index of the target link and the baseline anomaly judgment threshold is used as the anomaly tolerance increment of the target link. The sum of the anomaly tolerance increment of the target link and the baseline anomaly judgment threshold is used as the anomaly judgment threshold of the target link. Similarly, the anomaly judgment thresholds of each link are obtained.
[0039] Wherein, the anomaly detection threshold for link i is The calculation formula is as follows:
[0040]
[0041] In the formula, Indicates the baseline anomaly detection threshold. The value range can be set to The older and more worn the target PCR equipment, the better. The higher the value, the better. This represents the anomaly tolerance index for stage i.
[0042] The smaller the value of the anomaly tolerance index, the more important the corresponding link or the more concentrated the state distribution. The smaller the anomaly judgment threshold of that link, the stricter the judgment standard. Even a small deviation in the characteristics of the equipment state is considered an anomaly. Conversely, when the anomaly tolerance index is high, the judgment standard is relatively lenient. Only when the characteristics of the equipment state deviate significantly from the historical normal baseline will it be judged as an anomaly. By dynamically adjusting the anomaly judgment threshold through the anomaly tolerance index, the judgment standard of each link can adapt to the characteristics of different links, thereby effectively balancing the anomaly detection rate and the false alarm rate, and improving the detection accuracy of abnormal states of PCR equipment.
[0043] Step S4: Based on the Gaussian mixture model, construct an adaptive clustering model for each stage.
[0044] Because the physical processes and fluid characteristics of PCR equipment differ at each stage, the historical equipment state vectors of different stages under normal operation exhibit heterogeneous multimodal distribution characteristics in the feature space. For example, the washing stage may form two or three normal state clusters due to differences in reagent batches, while the elution stage usually only shows a single-peak distribution. Therefore, this step establishes a stage-adaptive clustering strategy, independently constructing a Gaussian mixture model for each stage to achieve accurate modeling of the specific normal state distribution of the stage, forming a stage-adaptive normal state benchmark.
[0045] Specifically, a Gaussian mixture model is used to model the probability distribution of each stage. Based on actual application scenarios and human experience, a range of values for the number of Gaussian distributions is set, and each value within the range is substituted into the Gaussian mixture model. For the Gaussian mixture model corresponding to any stage, after substituting a certain value of the number of Gaussian distributions, the Gaussian mixture model is fitted using the expectation-maximization algorithm based on the historical equipment state vector set of that stage. To avoid local optima, multiple random initializations are performed to obtain multiple initial models corresponding to the value of the number of Gaussian distributions. Based on the existing probability density function, the log-likelihood value of each initial model is calculated, and the initial model with the largest log-likelihood value is selected as the initial clustering model corresponding to the value of the number of Gaussian distributions in the corresponding stage. Similarly, the initial clustering model corresponding to each value of the number of Gaussian distributions in each stage is obtained. The optimal number of Gaussian distributions in each stage is determined using the Bayesian information criterion, thereby obtaining the adaptive clustering model for each stage.
[0046] The core principle of using the Gaussian Mixture Model (GMM) lies in its ability to adaptively fit probability distributions of arbitrary shapes through a combination of multiple Gaussian distributions, thereby accurately characterizing the complex distribution features of normal states in each stage. Simultaneously, the soft clustering mechanism based on the probability density function quantifies the normality of historical equipment state vectors, providing a mathematical foundation for subsequent anomaly detection. The range of values for the number of Gaussian distributions can be adjusted according to specific implementation conditions; for example, the lower limit can be set to 1, and the upper limit to 8. Furthermore, to avoid local optima, 5 to 20 random initializations can be performed. When determining the optimal number of Gaussian distributions using the Bayesian Information Criterion (BIC), the value of the Gaussian distribution corresponding to the minimum BIC value is selected as the optimal number of Gaussian distributions.
[0047] Furthermore, based on the adaptive clustering model of each stage and the existing probability density function calculation formula, the probability density of all historical equipment state vectors in the historical equipment state vector set of each stage can be calculated. The probability density value is mapped by the logarithmic function, and the mapped value is used as the log-likelihood value corresponding to the historical equipment state vector. The maximum and minimum values of the log-likelihood values of each stage are counted to obtain the historical maximum log-likelihood value and the historical minimum log-likelihood value of each stage, so as to facilitate the subsequent calculation of state determination indicators.
[0048] Step S5: Obtain real-time data from the target PCR device. After the current step is completed, calculate the status judgment index of the current step based on the corresponding adaptive clustering model.
[0049] Specifically, starting from the beginning of the current meat nucleic acid testing process, the operating parameters of the rotary pump ring drive motor in the target PCR equipment are collected synchronously in real time at the same sampling frequency. After the current step is completed, the current equipment state vector corresponding to the current step is constructed through the same steps. The current equipment state vector is input into the adaptive clustering model corresponding to the current step to obtain the log-likelihood value of the current step. The historical maximum log-likelihood value and the historical minimum log-likelihood value corresponding to the current step are extracted based on the historical equipment state vector set. The difference between the historical maximum log-likelihood value and the historical minimum log-likelihood value is used as the normal benchmark. The difference between the historical maximum log-likelihood value and the log-likelihood value of the current step is used as the abnormal deviation feature of the current step. The ratio between the abnormal deviation feature and the normal benchmark is used as the state judgment index of the current step.
[0050] The current state determination index is SC, and its calculation formula is as follows:
[0051]
[0052] In the formula, This represents the log-likelihood value of the current device state vector. This represents the historical maximum log-likelihood value corresponding to the current stage. This represents the historical minimum log-likelihood value corresponding to the current stage.
[0053] The log-likelihood value of the current device state vector is used to measure the "reasonableness" of the current device state vector under the adaptive clustering model corresponding to the current stage; The larger the value, the more the current device state vector conforms to the clustering characteristics of the corresponding adaptive clustering model, and the closer the operating parameters of the drive motor in the current stage are to the historical normal level of that stage; conversely, The smaller the value, the greater the possibility of an anomaly in the current process; the value of the status judgment index represents the degree of abnormality of the current process that exceeds the historical normal range, and can directly indicate the possibility of an abnormal state in the current process. The higher the value, the greater the possibility that the current process is in an abnormal operating state.
[0054] Step S6: Based on the status judgment indicators of the current stage and the corresponding anomaly judgment threshold, detect and judge the equipment status of the current stage.
[0055] Specifically, the status judgment index of the current stage is compared with the corresponding anomaly judgment threshold. When the status judgment index is less than the anomaly judgment threshold, the target PCR device is determined to be operating normally in the current stage. When the status judgment index is greater than or equal to the anomaly judgment threshold, the target PCR device is determined to be in an abnormal operating state in the current stage and an anomaly alarm is triggered.
[0056] When an abnormal alarm is triggered, the abnormal step label, timestamp, and corresponding current device status vector can be recorded in the log of the PCR equipment control system for reference during subsequent PCR equipment operation and maintenance.
[0057] This invention also discloses a device status detection system for PCR, used to implement the above-described device status detection method for PCR, the system structure of which is as follows: Figure 2 As shown, it includes: a processor, a memory, a communication interface, an early warning device, and a PCR equipment control system. The processor stores computer program instructions for implementing the above-mentioned method for detecting the status of a PCR device. The communication interface is communicatively connected to the PCR equipment control system and the early warning device.
[0058] Specifically, the early warning device can be a buzzer. When an abnormal operating state is detected in a certain stage of the PCR equipment, an alarm can be sounded through the buzzer, or the abnormal signal can be transmitted to the PCR equipment control system at the same time, and a visual alarm of the abnormal operating state can be displayed on the screen of the PCR equipment.
[0059] The embodiments included in this invention are merely descriptions of preferred embodiments of the invention and are not limited to the precise structures described above and shown in the accompanying drawings. Various modifications and changes can be made without departing from the scope of protection. Any variations and improvements made by those skilled in the art to the technical solutions of this invention without departing from the design concept of this invention should fall within the scope of protection of this invention.
Claims
1. A method for detecting the status of equipment used in PCR, characterized in that: The historical normal operating parameters of the drive motor under each stage are obtained, and the corresponding status data segment set for each stage is obtained; based on the status data segment set of each stage, the historical equipment status vector set of each stage is constructed; based on the historical equipment status vector set of each stage, the anomaly judgment threshold of each stage is calculated respectively. Based on the Gaussian mixture model, an adaptive clustering model is constructed for each step; real-time data of the target PCR equipment is acquired, and after the current step is completed, the status judgment index of the current step is calculated based on the corresponding adaptive clustering model; based on the status judgment index of the current step and the corresponding anomaly judgment threshold, the equipment status of the current step is detected and judged. Among them, based on several PCR devices of the same model as the target PCR device, in multiple identical historical normal meat nucleic acid testing processes, the historical normal operation parameters of the rotary pump ring drive motor under each step are collected synchronously at a fixed sampling frequency, including torque data, speed data, and current data, to obtain multiple three-dimensional historical normal sequences corresponding to each step. The maximum historical normal torque, maximum historical normal speed, and maximum historical normal current are extracted from all historical normal operating parameters. For any three-dimensional historical normal sequence in the state data segment set of a certain link, the ratio between the mean of each data point in the torque historical normal sequence and the maximum historical normal torque is taken as the first flow component; the ratio between the variance of each data point in the speed historical normal sequence and the square of the maximum historical normal speed is taken as the second flow component; the ratio between the mean of each data point in the current historical normal sequence and the maximum historical normal current is taken as the first system component; and the coefficient of variation of the current historical normal sequence is taken as the second system component. The four components corresponding to the same three-dimensional historical normal sequence are merged into a historical equipment state vector. Similarly, the historical equipment state vector set of each link is obtained. Gaussian mixture models (GMMs) are used to model the probability distribution of each stage. A range of values for the Gaussian distribution quantity is defined, and each value is substituted into the GMM. For any stage's GMM, a specific Gaussian distribution quantity is substituted into the model. Based on the historical equipment state vector set of that stage, the expectation-maximization algorithm is used to fit the GMM. To avoid local optima, multiple random initializations are performed, resulting in multiple initial models corresponding to that Gaussian distribution quantity. Based on the existing probability density function, the log-likelihood value of each initial model is calculated. The initial model with the largest log-likelihood value is selected as the initial clustering model corresponding to that Gaussian distribution quantity for that stage. Similarly, the initial clustering model corresponding to each Gaussian distribution quantity for each stage is obtained. The optimal Gaussian distribution quantity for each stage is determined using the Bayesian information criterion, thus obtaining the adaptive clustering model for each stage.
2. The method for detecting the status of equipment for PCR according to claim 1, characterized in that, The division to obtain the set of state data segments corresponding to each stage includes: obtaining the stage label and runtime of each stage from the control system of the PCR equipment, obtaining the stage label and runtime corresponding to each three-dimensional historical normal sequence, and calculating the total runtime of each historical normal meat nucleic acid detection process, numbering each stage in sequence according to the order of each stage, classifying the three-dimensional historical normal sequences according to the stage label, and taking the combination of the three-dimensional historical normal sequence and the corresponding stage label of each stage as the set of state data segments corresponding to each stage.
3. The method for detecting the status of equipment for PCR according to claim 1, characterized in that, The calculation of the anomaly judgment threshold for each stage includes: selecting any stage as the target stage, extracting the runtime corresponding to each three-dimensional historical normal sequence under the target stage, taking the average of the ratios between each runtime and the corresponding total runtime as the importance weight of the target stage, and taking the value of 1 minus the importance weight as the first sensitivity coefficient of the target stage. Similarly, the first sensitivity coefficient of each stage is obtained; extracting the historical equipment state vector set of the target stage, extracting the four components of each historical equipment state vector to form four component sequences, calculating the coefficient of variation of each of the four component sequences, taking the average of the four coefficients of variation as the anomaly tolerance level of the target stage, and similarly obtaining the anomaly tolerance level of each stage, then normalizing the anomaly tolerance level of each stage, taking the value of the normalized anomaly tolerance level as the second sensitivity coefficient of the corresponding stage, and taking the product between the first sensitivity coefficient and the corresponding second sensitivity coefficient of each stage as the anomaly tolerance index of each stage.
4. The method for detecting the status of equipment for PCR according to claim 3, characterized in that, The calculation of the anomaly judgment threshold for each stage also includes: setting a safety threshold for the anomaly tolerance index based on the detection accuracy requirements of the actual application scenario and human experience; when the value of the anomaly tolerance index of a certain stage exceeds the safety threshold, the safety threshold is used as the value of the anomaly tolerance index for that stage; setting a benchmark anomaly judgment threshold based on the life cycle and operating wear and tear of the target PCR equipment; using the product between the anomaly tolerance index of the target stage and the benchmark anomaly judgment threshold as the anomaly tolerance increment of the target stage; and using the sum of the anomaly tolerance increment of the target stage and the benchmark anomaly judgment threshold as the anomaly judgment threshold of the target stage. Similarly, the anomaly judgment thresholds for each stage are obtained.
5. The method for detecting the status of equipment for PCR according to claim 1, characterized in that, The calculation of the current stage's state determination index includes: starting from the beginning of the current meat nucleic acid testing process, synchronously collecting the operating parameters of the rotary pump ring drive motor in the target PCR equipment in real time at the same sampling frequency; after the current stage is completed, constructing the current equipment state vector corresponding to the current stage through the same steps; inputting the current equipment state vector into the adaptive clustering model corresponding to the current stage to obtain the log-likelihood value of the current stage; extracting the historical maximum log-likelihood value and historical minimum log-likelihood value corresponding to the current stage based on the historical equipment state vector set; using the difference between the historical maximum log-likelihood value and the historical minimum log-likelihood value as the normal benchmark; using the difference between the historical maximum log-likelihood value and the log-likelihood value of the current stage as the abnormal deviation feature of the current stage; and using the ratio between the abnormal deviation feature and the normal benchmark as the current stage's state determination index.
6. The method for detecting the status of equipment for PCR according to claim 5, characterized in that, The detection and judgment of the equipment status in the current stage includes: comparing the status judgment index of the current stage with the anomaly judgment threshold corresponding to the current stage; when the status judgment index is less than the anomaly judgment threshold, it is determined that the target PCR equipment is operating normally in the current stage; when the status judgment index is greater than or equal to the anomaly judgment threshold, it is determined that the target PCR equipment has an abnormal operating state in the current stage and an anomaly alarm is triggered.
7. A device status detection system for PCR, characterized in that: The device includes a processor, a memory, a communication interface, an early warning device, and a PCR equipment control system. The processor stores computer program instructions for implementing the device status detection method for PCR as described in any one of claims 1 to 6. The communication interface is communicatively connected to the PCR equipment control system and the early warning device.