Intermediate preparation perception method for piperacillin acid preparation process

Abstract

The present invention relates to the technical field of production data monitoring and management, and in particular to a method for sensing the preparation of an intermediate in a piperacillin acid preparation process. The method collects parameter data in a reaction vessel in different process steps. Different preparation processes are classified by constructing a short-term state mapping vector and statistically analyzing local temperature information to obtain a preparation state category. A prediction network is trained by analyzing the state transition confidence of each preparation process in a database, and then the state transition information of the real-time preparation process is predicted, and abnormal reaction vessels are sensed by real-time state transition features and previous state transition features to avoid abnormal intermediates in the abnormal reaction vessel from being mixed into a subsequent preparation process. The present invention avoids abnormal intermediates from being mixed into the piperacillin acid preparation process by monitoring and analyzing production data, thereby ensuring the yield rate of the product.

Description

Technical Field

[0001] The present invention relates to the technical field of production data monitoring and management, and in particular to an intermediate preparation sensing method used in the piperacillin acid preparation process. Background Art

[0002] Piperacillin acid is an important drug for treating bacterial infections. A variety of chemicals are required in the preparation process of piperacillin acid. N-ethyldioxypiperazinecarbonyl chloride is an important pharmaceutical intermediate. Its preparation directly affects the preparation quality of piperacillin acid. Therefore, the preparation process of the intermediate needs to be integrated in the preparation process of piperacillin acid, and the preparation quality of the intermediate needs to be perceived during the preparation of the intermediate. If the preparation quality of the intermediate cannot be guaranteed, it will directly affect the quality of the piperacillin acid product, making the preparation process substandard, and increasing the difficulty of later drug cleaning, increasing costs and causing cost waste. Summary of the invention

[0003] In order to solve the above technical problems, the object of the present invention is to provide a method for sensing the preparation of an intermediate in the preparation process of piperacillin acid, and the technical solution adopted is as follows:

[0004] The present invention proposes a method for sensing the preparation of an intermediate in the preparation process of piperacillin acid, the method comprising:

[0005] According to a preset sampling frequency, the overall temperature sequence of the reaction container and the local temperature sequences at different positions in the preset sampling period during the TMCS reaction process are obtained;

[0006] After the TMCS reaction process is completed, TEA is added into the reaction container, and the overall temperature in the reaction container is controlled within a preset temperature range by a temperature controller, and a local temperature critical value at each position is obtained, and a local temperature difference sequence of the local temperature critical value and the corresponding local temperature sequence is obtained;

[0007] After the TEA reaction is completed, the process is filtered and washed, and a diphosgene reaction process is performed. After the diphosgene reaction process is completed, an evaporation process is performed to obtain a weight sequence during the evaporation process;

[0008] Obtain a short-term state mapping vector according to the volatility of the elements in the overall temperature sequence and the difference of the elements in the weight sequence;

[0009] The data of different preparation processes are collected in the database, and the preparation processes are divided into multiple preparation state categories according to the first similarity of the local temperature difference sequence between different preparation processes and the second similarity of the short-time state mapping vector; the K nearest neighbor group of each target preparation process is obtained, and the state transition confidence set is obtained according to the preparation state category to which the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group belongs; the prediction network is trained according to the short-time state mapping vector of each preparation process and its corresponding state transition confidence set;

[0010] The short-term state mapping vector of the real-time preparation process is input into the prediction network to obtain a real-time confidence set; a reference confidence set of the previous preparation process of the reaction container corresponding to the real-time preparation process is obtained; a state sudden change index is obtained according to the difference between the real-time confidence set and the reference confidence set, and whether the corresponding reaction container is abnormal is sensed according to the state sudden change index. If an abnormality occurs, the intermediate prepared by the abnormal reaction container is not used in the subsequent preparation process of piperacillin acid.

[0011] Furthermore, the method for acquiring the local temperature sequence includes:

[0012] Eight temperature sensors are installed at the upper and lower positions of four equally divided sides of the reaction container to obtain local temperature sequences at different positions.

[0013] Furthermore, the method for obtaining the local temperature critical value includes:

[0014] When the temperature controller controls the overall temperature in the reaction container within a preset temperature range, the local temperature values ​​at different positions in the reaction container when the overall temperature reaches the highest temperature are used as local temperature critical values.

[0015] Further, the obtaining of the short-term state mapping vector according to the volatility of the overall temperature sequence elements and the difference of the elements in the weight sequence comprises:

[0016] The temperature element volatility is obtained according to the temperature element volatility formula, which includes:

[0017] T=ln[1+std(V)]

[0018] Where T is the volatility of the temperature element, std(V) is the standard deviation of the overall temperature sequence V;

[0019] The crystal weight difference is obtained according to the crystal weight difference formula, and the crystal weight difference formula includes:

[0020]

[0021] Where P is the crystal weight difference, OS maxis the maximum value in the weight sequence, OS min is the minimum value in the weight sequence, N is the number of elements in the weight sequence, OS i is the value of the i-th element in the weight sequence;

[0022] Taking the temperature element fluctuation as the starting point and the crystal weight difference as the end point, the short-time state mapping vector is obtained.

[0023] Furthermore, the method of dividing the preparation processes into a plurality of preparation state categories according to the first similarity of the local temperature difference sequences between different preparation processes and the second similarity of the short-time state mapping vectors includes:

[0024] Obtain the dynamic time warping distance of the local temperature difference sequence between corresponding positions between different preparation processes, use the negative number of the dynamic time warping distance as the power of an exponential function with a natural constant as the base, obtain the initial first similarity between different positions, and use the maximum initial first similarity as the first similarity between different preparation processes;

[0025] The cosine similarity between short-term state mapping vectors is used as the second similarity;

[0026] The product of the first similarity and the second similarity is used as the correlation between different preparation processes. The different preparation processes are clustered according to the correlation using a density clustering algorithm to obtain multiple preparation state categories.

[0027] Furthermore, the step of obtaining a state transition confidence set according to the preparation state category to which the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group belongs includes:

[0028] The preparation state category of the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group is counted, the number of samples of each preparation state category that has a corresponding relationship in the K nearest neighbor group is normalized, the state transition confidence corresponding to the preparation state category is obtained, and the state transition confidence set of the target preparation process corresponding to the K nearest neighbor group is obtained.

[0029] Further, obtaining the state sudden change index according to the difference between the real-time confidence set and the reference confidence set includes:

[0030] Respectively remove elements whose values ​​are less than a preset confidence threshold from the real-time confidence set and the reference confidence set to obtain a first comparison real-time confidence set and a first comparison reference confidence set;

[0031] Retain the elements in the first comparison real-time confidence set that are smaller than the element mean, and obtain a second comparison real-time confidence set;

[0032] Retain the elements in the first comparison reference confidence set that are greater than the element mean, and obtain a second comparison reference confidence set;

[0033] The intersection-and-joint ratio of the second comparison real-time confidence set and the second comparison reference confidence set is used as the state sudden change indicator.

[0034] Further, sensing whether the corresponding reaction container is abnormal according to the state sudden change indicator includes:

[0035] If the state sudden change index of the real-time preparation process of the target reaction container for a preset number of consecutive times is greater than the preset state sudden change index threshold, the target reaction container is considered to be a suspected abnormal container;

[0036] The preparation state category corresponding to the maximum element in the state transfer confidence of the real-time preparation process is taken as the transfer preparation state category, and the category difference between the transfer preparation state category and the preparation state category to which the current real-time preparation process belongs is counted; if the average category difference of the real-time preparation processes of the preset number of suspected abnormal containers is greater than the preset category difference threshold, the suspected abnormal container is considered to be an abnormal reaction container.

[0037] The present invention has the following beneficial effects:

[0038] The embodiment of the present invention collects a variety of data in the reaction container during the preparation process of the intermediate of piperacillin acid, takes into account the volatility of temperature data during the reaction process and the formation speed of crystal quality during evaporation and crystallization to construct a short-term state mapping vector, and uses the short-term state mapping vector to characterize the starting state and the ending state in the entire intermediate preparation process, providing an effective reference for the subsequent preparation state classification. Different preparation processes are classified by local temperature difference sequence and short-term state mapping vector, and the state transition confidence of each preparation process corresponding to each reaction container is analyzed, and then a prediction network is trained. The prediction network is used to obtain the confidence set of the real-time preparation process and the previous preparation process, and the corresponding reaction container is judged by comparing the state transition confidence of two consecutive preparation processes. Whether an abnormality occurs, thereby avoiding the intermediate in the abnormal reaction container from affecting the subsequent preparation of piperacillin acid, ensuring the effect and quality of the piperacillin acid preparation process, avoiding cost waste, and improving the yield rate of the prepared product. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1A flow chart of a method for sensing the preparation of an intermediate in the preparation process of piperacillin acid provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method for preparing an intermediate used in the preparation process of piperacillin acid according to the present invention, its specific implementation method, structure, characteristics and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The scenario targeted by the embodiment of the present invention is a piperacillin acid preparation process integrated with an N-ethyldioxypiperazinecarbonyl chloride preparation process. The specific preparation process includes:

[0044] 1. EDP (N-ethyldioxypiperazine) 142.3 g, MC 1400 ml, cooled to 0°C;

[0045] 2. Slowly add 168 ml of TMCS (trimethylchlorosilane) at 0°C, react at 0°C for 1 hour, and cool to <-20°C;

[0046] 3. Add 184 ml of TEA (triethylamine), add dropwise at <-15°C, and react at -5 to -10°C for 1 hour;

[0047] 4. Remove TEA hydrochloride by suction filtration and wash with a small amount of MC (dichloromethane);

[0048] 5. Cool down to -20℃, add 72ml of diphosgene dropwise, and react at room temperature for 1 hour;

[0049] 6. Evaporate to dryness using a reaction evaporator, add 400 ml of EA (ethyl acetate) to obtain N-ethyldioxypiperazinecarbonyl chloride for later use.

[0050] 7. Add 50g of ampicillin trihydrate, 25g of sodium bicarbonate, 400ml of water and 350ml of EA into a 1000ml reaction bottle;

[0051] 8. Slowly add about 63 ml of N-ethyldioxypiperazinecarbonyl chloride, adjust the pH to 6.4, and react for 10 minutes;

[0052] 9. Adjust the pH to 1.5-2.0 with concentrated hydrochloric acid and grow the crystals at 0-5℃ for 5-6 hours;

[0053] 10. Filter with suction, wash with a small amount of water, and react at less than 50°C to obtain about 60 g of piperacillin acid.

[0054] The following is a detailed description of a specific scheme of the intermediate preparation sensing method for the preparation process of piperacillin acid provided by the present invention in conjunction with the accompanying drawings.

[0055] See also Figure 1 , which shows a flow chart of an intermediate preparation sensing method for piperacillin acid preparation provided by one embodiment of the present invention, the method comprising:

[0056] Step S1: obtaining the overall temperature sequence of the reaction container and the local temperature sequences at different positions within a preset sampling period during the TMCS reaction process according to a preset sampling frequency.

[0057] The embodiment of the present invention aims to monitor the parameter data of the reaction container during the intermediate preparation process to achieve perception of the intermediate preparation. Therefore, in the specific preparation process targeted by the embodiment of the present invention, it is necessary to monitor the parameter data in process steps 1 to 6.

[0058] Because the reaction container needs to ensure the temperature inside the container in process step 2 of the preparation process, that is, to ensure that the TMCS reacts in a 0° environment, it is necessary to monitor the temperature data inside the reaction container during the TMCS reaction, and obtain the overall temperature sequence of the reaction container and the local temperature sequence at different positions according to the preset sampling frequency and the preset sampling period. The overall temperature reflects the temperature control ability of the reaction container during the TMCS reaction process. The smaller the temperature fluctuation in the time series, the higher the temperature fluctuation, the better the temperature control ability of the reaction container. The local temperature sequence at different positions can reflect the changes in the internal convection and insulation state of the container during the working process.

[0059] The methods for specific activities of local temperature sequence include:

[0060] Eight temperature sensors are installed at the upper and lower positions of four equally divided sides of the reaction container, that is, each side has temperature information of the upper and lower parts, and the local temperature sequence of different positions is obtained.

[0061] It should be noted that the sampling frequency and sampling period need to be set according to the specific implementation scenario and sensor capabilities, and are not limited here.

[0062] Step S2: After the TMCS reaction process is completed, TEA is added into the reaction container, and the overall temperature in the reaction container is controlled within a preset temperature range by a temperature controller, and the local temperature critical value of each position is obtained, and the local temperature critical value and the local temperature difference sequence corresponding to the local temperature sequence are obtained.

[0063] After the TMCS reaction is completed, process step 3 in the preparation process is performed, that is, TEA is added to the reaction vessel. It can be seen from the content of process step 3 that the process of adding TEA is carried out at a temperature less than -15°, and the reaction process is carried out at a temperature of -5° to -10°. Therefore, the temperature control requirements for the reaction vessel during the TEA reaction are more stringent, and the overall temperature in the reaction vessel needs to be controlled within a preset temperature range by a temperature controller. In order to ensure that the temperature in the reaction vessel meets the process requirements, the temperature in the container needs to be increased, but due to temperature inertia and the PID parameters of the temperature controller in the reaction vessel, the temperature continues to increase after it reaches the target value that meets the preset temperature range. Therefore, the entire TEA reaction process is observed. In the process of the temperature controller controlling the overall temperature in the reaction vessel within the preset temperature range, the local temperature values ​​at different positions in the reaction vessel when the overall temperature reaches the highest temperature are used as the local temperature critical value. The local temperature critical value can represent the reference temperature when trying to reduce or maintain the temperature after the temperature overshoot or stabilization caused by the temperature control system increasing the temperature during the TEA reaction. Therefore, the local temperature critical value and the local temperature difference sequence of the corresponding local temperature sequence are obtained, that is, the absolute value of the difference between the local temperature critical value and the corresponding local temperature value is used as the element in the local temperature difference sequence. The local difference sequence can be used to characterize whether the temperature at different locations in the temperature control process meets the change characteristics. If the element value in a local temperature difference sequence is large, it means that the difference between the local temperature and the local temperature critical value is larger, which is easy to cause instability in temperature control, that is, the reaction vessel is more likely to be abnormal.

[0064] Step S3: After the TEA reaction is completed, suction filtration and washing are performed and a diphosgene reaction process is performed. After the diphosgene reaction process is completed, an evaporation process is performed to obtain a weight sequence during the evaporation process.

[0065] The subsequent process steps of the preparation process are performed. From the specific content of the preparation process, it can be seen that process steps 4 to 6 do not involve long-term complex temperature control of the reaction vessel. Therefore, only the last preparation process in the intermediate preparation process, that is, the evaporation process, is analyzed. After the TEA reaction is completed, suction filtration and washing are performed and a diphosgene reaction process is performed. After the diphosgene reaction process is completed, an evaporation process is performed to obtain a weight sequence in the evaporation process. The evaporation process is the crystallization process of the intermediate. The speed of crystallization and the weight of the crystallization directly reflect the efficiency of the reaction vessel controlling the reaction evaporator. Therefore, the weight sequence can represent the state at the end of the intermediate preparation process.

[0066] Step S4: Obtain a short-term state mapping vector according to the volatility of the overall temperature sequence elements and the difference of the elements in the weight sequence.

[0067] From the data collection period and content definition of the overall temperature sequence and weight sequence, it can be seen that the overall temperature sequence represents the early preparation state of the intermediate preparation process, and the weight sequence represents the final preparation state of the intermediate preparation process. Therefore, the short-term state mapping vector can be obtained according to the element volatility of the overall temperature sequence and the difference of the elements in the weight sequence. The short-term state mapping vector represents the state characteristics of the reaction container during the entire intermediate preparation process. The higher the element volatility of the overall temperature sequence, the more abnormal the temperature control of the reaction container; the greater the difference of the elements in the weight sequence, the higher the crystallization efficiency of the reaction container during the crystallization process. The specific acquisition method includes:

[0068] The temperature element volatility is obtained according to the temperature element volatility formula, which includes:

[0069] T=ln[1+std(V)]

[0070] Among them, T is the temperature element volatility, and std(V) is the standard deviation of the overall temperature sequence V. That is, the larger the standard deviation, the stronger the element volatility, the worse the temperature control ability of the reaction container, and the more abnormal the container state.

[0071] The crystal weight difference is obtained according to the crystal weight difference formula, and the crystal weight difference formula includes:

[0072]

[0073] Where P is the crystal weight difference, OS max is the maximum value in the weight sequence, OS min is the minimum value in the weight sequence, N is the number of elements in the weight sequence, OS i is the value of the ith element in the weight sequence. In the formula for the crystallization weight difference, the element mean is used as the denominator for the purpose of normalization; the range is used as the numerator. The larger the range, the higher the efficiency of the crystallization process.

[0074] Taking the temperature element fluctuation as the starting point and the crystal weight difference as the end point, the short-time state mapping vector is obtained.

[0075] Step S5: Count the data of different preparation processes in the database, and divide the preparation processes into multiple preparation state categories according to the first similarity of the local temperature difference sequence between different preparation processes and the second similarity of the short-time state mapping vector; obtain the K nearest neighbor group of each target preparation process, and obtain the state transition confidence set according to the preparation state category to which the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group belongs; train the prediction network according to the short-time state mapping vector of each preparation process and its corresponding state transition confidence set.

[0076] The database contains data of multiple intermediate preparation processes, and the preparation processes include corresponding reaction container labels and preparation time. Data of different preparation processes are counted in the database, and the preparation processes are divided into multiple preparation state categories according to the first similarity of the local temperature difference sequence between different preparation processes and the second similarity of the short-time state mapping vector, that is, the preparation states in the same preparation state category are in similar preparation states, and different preparation state categories represent different preparation states.

[0077] In an embodiment of the present invention, after obtaining the preparation status categories, the preparation status categories are sorted according to the mean value of the crystal weight finally obtained in the preparation process in each preparation status category. The larger the mean value of the crystal weight, the higher the ranking of the preparation status category, and the higher the ranking, the more normal the corresponding preparation status.

[0078] The specific preparation status classification process includes:

[0079] The dynamic time warping distance of the local temperature difference sequence between corresponding positions in different preparation processes is obtained, and the negative number of the dynamic time warping distance is used as the power of an exponential function with a natural constant as the base to obtain the initial first similarity between different positions, that is, the expression is: Where e is a natural constant, is the local temperature difference series and the local temperature difference series The dynamic time warping distance between them.

[0080] The maximum initial first similarity is taken as the first similarity between different preparation processes.

[0081] The cosine similarity between short-term state mapping vectors is used as the second similarity.

[0082] The product of the first similarity and the second similarity is used as the correlation between different preparation processes. The different preparation processes are clustered according to the correlation using a density clustering algorithm to obtain multiple preparation state categories.

[0083] It should be noted that the density clustering algorithm and the dynamic time warping distance algorithm are well known to those skilled in the art and will not be described in detail here. The number of preparation state categories can be set according to the specific implementation scenario and is set to 5 in the embodiment of the present invention.

[0084] Further obtain the K nearest neighbor group of the target preparation process, each preparation process sample in the K nearest neighbor group is a sample with a high correlation with the target preparation process. It should be noted that the method for obtaining the K nearest neighbor group is a technical means well known to those skilled in the art and will not be described in detail here. The state transition confidence set is obtained according to the preparation state category to which the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group belongs, specifically including:

[0085] The preparation state category of the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group is counted, and the number of samples with corresponding relationships in each preparation state category in the K nearest neighbor group is normalized to obtain the state transition confidence corresponding to the preparation state category, and the state transition confidence set of the target preparation process corresponding to the K nearest neighbor group is obtained. For example, there are 10 samples in the K nearest neighbor group, among which there are 5 samples of the preparation state category A corresponding to the next preparation process in the time series, then the corresponding state transition confidence can be obtained by normalizing 5.

[0086] In the embodiment of the present invention, the number of samples in the K nearest neighbor group is set to 10, and the normalization method adopts range normalization.

[0087] The prediction network is trained according to the short-time state mapping vector of each preparation process and its corresponding transfer confidence set. That is, in the prediction network, the input data is the short-time state mapping vector, and the output is the corresponding state transfer confidence. The preparation state category corresponding to the next preparation process can be determined according to the output state transfer confidence, that is, the preparation state category corresponding to the maximum element in the state transfer confidence is the preparation state category corresponding to the next preparation process.

[0088] In the embodiment of the present invention, the prediction network adopts a fully convolutional neural network, the network task is a classification task, and the specific network structure and training method are technical means well known to those skilled in the art, which will not be elaborated here.

[0089] Step S6: input the short-term state mapping vector of the real-time preparation process into the prediction network to obtain a real-time confidence set; obtain a reference confidence set of the previous preparation process of the reaction container corresponding to the real-time preparation process; obtain a state sudden change index according to the difference between the real-time confidence set and the reference confidence set, and sense whether the corresponding reaction container is abnormal according to the state sudden change index. If an abnormality occurs, the intermediate prepared by the abnormal reaction container is not used in the subsequent preparation process of piperacillin acid.

[0090] The previous preparation process of the corresponding reaction container of the real-time preparation process is used as a reference, and the real-time confidence set and the reference confidence set are obtained through the prediction network. The real-time confidence set is the state transition feature of the current real-time preparation process at the next moment, and the reference confidence is the transfer feature of the previous preparation process to the real-time preparation process. If the two transfer features are very different, it means that there is a sudden change in the preparation state of the current reaction container, and the reaction process of the reaction container is unstable. Therefore, the state sudden change index can be obtained according to the difference between the real-time confidence set and the reference confidence set, and the state sudden change index can be used to perceive whether the corresponding reaction container is abnormal. If an abnormality occurs, the intermediate prepared by the abnormal reaction container is not used in the subsequent preparation of piperacillin acid, so as to avoid the low-quality intermediates in the abnormal reaction container affecting the quality of the subsequent piperacillin acid preparation product.

[0091] Specifically, the state sudden change index is obtained according to the difference between the real-time confidence set and the reference confidence set, including:

[0092] The elements in the real-time confidence set and the reference confidence set that are less than the preset confidence threshold are removed respectively, with the purpose of deleting meaningless low-confidence preparation state categories to avoid the influence of such data on the state transition judgment, and obtaining the first comparison real-time confidence set and the first comparison reference confidence set. It should be noted that in the embodiment of the present invention, the position of each element in the state transition confidence set represents a preparation state category of different levels, the element value is the corresponding state transition confidence, and the deletion operation refers to setting the element value of the corresponding position to 0.

[0093] The elements in the first comparison real-time confidence set that are less than the element mean are retained to obtain the second comparison real-time confidence set. The elements in the first comparison reference confidence set that are greater than the element mean are retained to obtain the second comparison reference confidence set. The intersection and union ratio of the second comparison real-time confidence set and the second comparison reference confidence set is used as the state sudden change index. The larger the state sudden change index is, the more obvious the overlap between the last low confidence situation and the current high confidence situation is. Therefore, the state transition relationship when the reaction container is working is not common, and the more abnormal the reaction container is.

[0094] Specifically, the following are used to detect whether the corresponding reaction container is abnormal based on the state sudden change indicator:

[0095] If the state sudden change index of the real-time preparation process of the target reaction container for a preset number of consecutive times is greater than the preset state sudden change index threshold, the target reaction container is considered to be a suspected abnormal container.

[0096] The preparation state category corresponding to the maximum element in the state transfer confidence of the real-time preparation process is taken as the transfer preparation state category, and the category difference between the transfer preparation state category and the preparation state category to which the current real-time preparation process belongs is counted; if the average category difference of the real-time preparation processes of the preset number of suspected abnormal containers is greater than the preset category difference threshold, the suspected abnormal container is considered to be an abnormal reaction container. In the embodiment of the present invention, the category difference refers to the level difference of the preparation state category. For example, the preparation state category corresponding to the current preparation process is level 2, and the preparation state category corresponding to the next preparation process is level 8, then the category difference is 6.

[0097] By sensing and judging abnormal reaction vessels, low-quality intermediates can be prevented from mixing into the subsequent preparation process of piperacillin acid, thereby ensuring the yield and preparation efficiency of piperacillin acid products.

[0098] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0099] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for sensing the preparation of an intermediate in the preparation process of piperacillin acid, characterized in that: The method comprises: According to a preset sampling frequency, the overall temperature sequence of the reaction container and the local temperature sequences at different positions in the preset sampling period during the TMCS reaction process are obtained; After the TMCS reaction process is completed, TEA is added into the reaction container, and the overall temperature in the reaction container is controlled within a preset temperature range by a temperature controller, and a local temperature critical value at each position is obtained, and a local temperature difference sequence of the local temperature critical value and the corresponding local temperature sequence is obtained; After the TEA reaction is completed, the process is filtered and washed, and a diphosgene reaction process is performed. After the diphosgene reaction process is completed, an evaporation process is performed to obtain a weight sequence during the evaporation process; Obtain a short-term state mapping vector according to the volatility of the elements in the overall temperature sequence and the difference of the elements in the weight sequence; The data of different preparation processes are collected in the database, and the preparation processes are divided into multiple preparation state categories according to the first similarity of the local temperature difference sequence between different preparation processes and the second similarity of the short-time state mapping vector; the K nearest neighbor group of each target preparation process is obtained, and the state transition confidence set is obtained according to the preparation state category to which the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group belongs; the prediction network is trained according to the short-time state mapping vector of each preparation process and its corresponding state transition confidence set; Inputting the short-term state mapping vector of the real-time preparation process into the prediction network to obtain a real-time confidence set; obtaining a reference confidence set of the previous preparation process of the reaction container corresponding to the real-time preparation process; obtaining a state sudden change index according to the difference between the real-time confidence set and the reference confidence set, and sensing whether the corresponding reaction container is abnormal according to the state sudden change index. If an abnormality occurs, the intermediate prepared by the abnormal reaction container is not used in the subsequent preparation process of piperacillin acid; The step of obtaining a short-term state mapping vector according to the volatility of the overall temperature sequence elements and the difference of the elements in the weight sequence comprises: The temperature element volatility is obtained according to the temperature element volatility formula, which includes: in, is the temperature element volatility, The overall temperature series The standard deviation of The crystal weight difference is obtained according to the crystal weight difference formula, and the crystal weight difference formula includes: in, is the crystal weight difference, is the maximum value in the weight sequence, is the minimum value in the weight sequence, is the number of elements in the weight sequence, is the first element value; Taking the temperature element fluctuation as the starting point and the crystal weight difference as the end point, the short-time state mapping vector is obtained.

2. The method for sensing the preparation of an intermediate used in the preparation process of piperacillin acid according to claim 1, characterized in that: The method for obtaining the local temperature sequence includes: Eight temperature sensors are installed at the upper and lower positions of four equally divided sides of the reaction container to obtain local temperature sequences at different positions.

3. The method for sensing the preparation of an intermediate used in the preparation process of piperacillin acid according to claim 1, characterized in that: The method for obtaining the local temperature critical value includes: When the temperature controller controls the overall temperature in the reaction container within a preset temperature range, the local temperature values ​​at different positions in the reaction container when the overall temperature reaches the highest temperature are used as local temperature critical values.

4. The method for sensing the preparation of an intermediate used in the preparation process of piperacillin acid according to claim 1, characterized in that: The method of dividing the preparation processes into a plurality of preparation state categories according to the first similarity of the local temperature difference sequences between different preparation processes and the second similarity of the short-time state mapping vectors comprises: Obtain the dynamic time warping distance of the local temperature difference sequence between corresponding positions between different preparation processes, use the negative number of the dynamic time warping distance as the power of an exponential function with a natural constant as the base, obtain the initial first similarity between different positions, and use the maximum initial first similarity as the first similarity between different preparation processes; The cosine similarity between short-term state mapping vectors is used as the second similarity; The product of the first similarity and the second similarity is used as the correlation between different preparation processes. The different preparation processes are clustered according to the correlation using a density clustering algorithm to obtain multiple preparation state categories.

5. The method for sensing the preparation of an intermediate used in the preparation process of piperacillin acid according to claim 1, characterized in that: The step of obtaining a state transition confidence set according to the preparation state category to which the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group belongs includes: The preparation state category of the next preparation process of the reaction container corresponding to each preparation process in the K nearest neighbor group is counted, the number of samples of each preparation state category that has a corresponding relationship in the K nearest neighbor group is normalized, the state transition confidence corresponding to the preparation state category is obtained, and the state transition confidence set of the target preparation process corresponding to the K nearest neighbor group is obtained.

6. The method for sensing the preparation of an intermediate used in the preparation process of piperacillin acid according to claim 1, characterized in that: The step of obtaining a state sudden change indicator according to a difference between the real-time confidence set and the reference confidence set includes: Respectively remove elements whose values ​​are less than a preset confidence threshold from the real-time confidence set and the reference confidence set to obtain a first comparison real-time confidence set and a first comparison reference confidence set; Retain the elements in the first comparison real-time confidence set that are smaller than the element mean, and obtain a second comparison real-time confidence set; Retain the elements in the first comparison reference confidence set that are greater than the element mean, and obtain a second comparison reference confidence set; The intersection-and-joint ratio of the second comparison real-time confidence set and the second comparison reference confidence set is used as the state sudden change indicator.

7. The method for sensing the preparation of an intermediate used in the preparation process of piperacillin acid according to claim 1, characterized in that: The sensing of whether the corresponding reaction container is abnormal according to the state sudden change indicator includes: If the state sudden change index of the real-time preparation process of the target reaction container for a preset number of consecutive times is greater than the preset state sudden change index threshold, the target reaction container is considered to be a suspected abnormal container; The preparation state category corresponding to the maximum element in the state transfer confidence of the real-time preparation process is taken as the transfer preparation state category, and the category difference between the transfer preparation state category and the preparation state category to which the current real-time preparation process belongs is counted; if the average category difference of the real-time preparation processes of the preset number of suspected abnormal containers is greater than the preset category difference threshold, the suspected abnormal container is considered to be an abnormal reaction container.

Images