Boiler water regulator production quality real-time monitoring method and system

Abstract

The invention relates to the technical field of industrial process control, in particular to a boiler water regulator production quality real-time monitoring method and system. According to the method, the abnormal evolution characteristics are extracted by constructing the state transition sequence, the abnormal states are sorted in combination with the grading priority, the influence degree of different state evolution on the production quality can be accurately recognized, meanwhile, the liquid level disturbance data in the corresponding abnormal response period are collected, and disturbance recovery time delay is calculated; the effectiveness difference of the abnormal state to the system response can be quantified, the offset situation is further identified based on the consistency between the scoring trend and the disturbance trend, the dynamic change rule of the response efficiency of the regulator is reversely deduced, the collaborative analysis of concentration and flow velocity change parameters is combined, and the accuracy of the system response efficiency is improved. According to the method, the whole-process associated monitoring of the quality change type and the reaction efficiency is realized, the process forms a nested linkage mechanism in multiple dimensions such as state structure identification, trend corresponding analysis and efficiency change judgment, and the potential quality shift phenomenon can be effectively identified.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial process control, and in particular to a method and system for real-time monitoring of the production quality of a boiler water conditioner. Background Art

[0002] The field of industrial process control technology involves the orderly regulation of material conversion and energy transfer processes in continuous or batch production processes, including the identification of the internal state of the production system, the analysis of key process variables, the control of the target product generation path, and the coordination and optimization of the overall production process.

[0003] Among them, the traditional real-time monitoring method of the production quality of boiler water conditioners refers to a method of continuously monitoring the quality status of the conditioner itself during the production stage when the conditioner is used in boiler water treatment to inhibit corrosion and deposition.

[0004] Traditional technologies mainly rely on simple measurements of physical or chemical parameters in the regulator production process. Their real-time monitoring capabilities are limited to numerical data collection at fixed time points. They lack the ability to characterize the continuity of state evolution and multi-dimensional linkage characteristics. When faced with the cross-evolution of multi-source states, it is difficult to effectively identify potential abnormal trajectories. In particular, in scenarios where temperature anomalies or flow rate deviations occur alternately during the production process, it is impossible to determine the correspondence between the state switching frequency and the system disturbance response, resulting in delayed response to abnormal phenomena and insufficient explanatory power. For example, when the liquid level system slowly recovers after the regulator addition operation, traditional methods cannot causally associate this phenomenon with the previous abnormal state, thereby creating the risk of misjudgment or omission, which is not conducive to effective tracking and tracing of quality status trends. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for real-time monitoring of the production quality of boiler water conditioner.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for real-time monitoring of the production quality of a boiler water conditioner, comprising the following steps:

[0007] S1: Collect state data from the boiler water conditioner production process and construct a transition sequence between states. Extract the evolution characteristics of each abnormal state in the transition sequence to obtain an abnormal feature dataset.

[0008] S2: Prioritize the evolution characteristics of each type of abnormal state in the abnormal feature dataset to obtain a state priority sequence;

[0009] S3: extracting abnormal state items of a specified priority in the state priority sequence, collecting liquid level disturbance data after the addition of boiler water conditioner during the time period when the abnormal state occurs, calculating the time delay of the corresponding liquid level disturbance recovery, and constructing a liquid level disturbance delay data set;

[0010] S4: Analyze the consistency between the state priority sequence and the score change trend and the disturbance recovery time trend under the same abnormal state item of the liquid level disturbance delay data set, determine the synchronization offset characteristics of the response period, and construct a response synchronization offset set;

[0011] S5: judging a change pattern of the regulator reaction efficiency according to the response synchronization offset set, and obtaining a monitoring result of the boiler water regulator quality status.

[0012] The improvements of the present invention are that the abnormal feature data set includes the evolution path length, state switching density, and transfer direction stability; the state priority sequence is specifically a scoring sequence, a priority label, and a sorting rank; the liquid level disturbance delay data set includes the disturbance start and end time, the liquid level change trend, and the disturbance recovery section; the response synchronization offset set is specifically a trend difference type, an offset state label, and a time period index; and the boiler water conditioner quality state monitoring result includes a reaction efficiency label, a state change type, and an abnormal response level.

[0013] The present invention is improved in that the specific steps of S1 are:

[0014] S101: Collect state data on concentration fluctuations, temperature deviations, and flow rate anomalies during the production process of boiler water conditioner, record the continuous residence time of each state on the time axis and the sequence of adjacent states, arrange the state flow in chronological order, identify state transition relationships, and obtain a continuous state transition sequence;

[0015] S102: Calculating the state transition probability distribution and the average dwell time between states in the continuous state transition sequence using a Markov chain model, identifying abnormal state segments of the continuous path characteristics, and extracting the switching frequency, transition direction repeatability, and dwell time combination indicators corresponding to the segments to obtain an abnormal state evolution structure feature group;

[0016] S103: Constructing a three-dimensional feature set with the state type as the index based on the abnormal state evolution structure feature group to obtain an abnormal feature data set.

[0017] The present invention is improved in that the specific steps of S2 are:

[0018] S201: Obtain the switching frequency, transfer direction repeatability, and dwell time combination index corresponding to the abnormal state in the abnormal feature data set, perform interval normalization processing on each feature item, and obtain a standardized evolution feature set;

[0019] S202: Based on the standardized evolution feature set, a weight coefficient of each combination indicator is set, and a priority score of each type of abnormal state is calculated by a weighted hierarchy function to obtain an abnormal state score result;

[0020] S203: According to the abnormal state scoring result, each abnormal state is sorted from high to low according to the score to obtain a state priority sequence.

[0021] The present invention is improved in that the specific steps of S3 are:

[0022] S301: extracting the abnormal state number of the specified order in the state priority sequence, searching for the corresponding abnormality occurrence time period, obtaining the start and end time points of the boiler water conditioner addition operation within the target time period, determining the positioning window for the addition response, and obtaining the response time interval corresponding to the abnormal state;

[0023] S302: Based on the response time interval corresponding to the abnormal state, collect the height change sequence recorded by the liquid level sensor in the corresponding time period, calibrate the time node between the disturbance start position and the first stable platform position, and obtain disturbance recovery response section data;

[0024] S303: Calculate the time interval for the liquid level to fall from the disturbance peak to the plateau state based on the disturbance recovery response segment data, establish a corresponding disturbance recovery delay according to the state number, and obtain a liquid level disturbance delay data set.

[0025] The present invention is improved in that the specific steps of S4 are:

[0026] S401: Based on the state priority sequence and the abnormal state number in the liquid level disturbance delay data set, extract the score and disturbance recovery time under the corresponding abnormal state item, and pair them in the order of the numbers to construct a state score and disturbance delay pairing group;

[0027] S402: Based on the state score and disturbance delay pairing group, sequentially compare the score change trend with the disturbance recovery time change trend, identify whether the change directions of adjacent abnormal state items in the two trend dimensions are consistent, record the state numbers where the trend directions are opposite or inconsistent, and generate a trend consistency offset identification set;

[0028] S403: According to the trend consistency offset identification set, mark the time period information corresponding to the abnormal state item where the offset occurs, summarize the state number, offset type and corresponding time index, and generate a response synchronization offset set.

[0029] The present invention is improved in that the specific steps of S5 are:

[0030] S501: calling the time period index in the response synchronization offset set, matching the abnormal state item with the corresponding number, extracting the concentration change and flow rate change rate corresponding to the target time period in the abnormal state evolution structure feature group, and generating a state association feature comparison table;

[0031] S502: Based on the state-related feature comparison table, perform combination relationship matching on each set of concentration change values ​​and flow rate change rates, determine the synergy and deviation degree of parameter fluctuations, classify them as reaction acceleration type, reaction lag type, or fluctuation indeterminate type, and generate a reaction efficiency change pattern label group;

[0032] S503: According to the reaction efficiency change pattern tag group, the change type, time period index and offset characteristics of each abnormal state item are integrated to establish a state-level quality monitoring tag and generate a boiler water conditioner quality state monitoring result.

[0033] A real-time monitoring system for the production quality of a boiler water conditioner, the system comprising:

[0034] Abnormal evolution extraction module: This module collects state data from the boiler water conditioner production process and constructs a transition sequence between states. It then extracts the evolutionary features of each abnormal state in the transition sequence to obtain an abnormal feature dataset.

[0035] State priority evaluation module: Priority scoring is performed on the evolution characteristics of each type of abnormal state in the abnormal feature data set to obtain a state priority sequence;

[0036] Disturbance delay analysis module: extracts abnormal state items of specified priority in the state priority sequence, collects liquid level disturbance data after the addition of boiler water conditioner during the time period when the abnormal state occurs, calculates the time delay of the corresponding liquid level disturbance response, and constructs a liquid level disturbance delay data set;

[0037] Response consistency judgment module: analyzes the consistency between the state priority sequence and the score change trend and the disturbance recovery time trend under the same abnormal state item in the liquid level disturbance delay dataset, determines the synchronization offset characteristics of the response period, and constructs a response synchronization offset set;

[0038] Efficiency change identification module: determines the change pattern of the regulator reaction efficiency according to the response synchronization offset set, and obtains the boiler water regulator quality status monitoring result.

[0039] Compared with the prior art, the advantages and positive effects of the present invention are:

[0040] In the present invention, by constructing a state transition sequence to extract abnormal evolution characteristics and sorting the abnormal states in combination with the scoring priority, the degree of influence of different state evolutions on production quality can be accurately identified. At the same time, the liquid level disturbance data during the corresponding abnormal response period is collected and the disturbance recovery time delay is calculated. The effectiveness difference of the abnormal state on the system response can be quantified. The offset situation is further identified based on the consistency between the scoring trend and the disturbance trend, thereby inferring the dynamic change law of the regulator response efficiency. Combined with the collaborative analysis of concentration and flow rate change parameters, the full-process correlation monitoring of quality change type and reaction efficiency is realized. This process forms a nested linkage mechanism in multiple dimensions such as state structure identification, trend correspondence analysis and efficiency change judgment. Compared with the method of only monitoring the physical properties of the regulator, it can effectively identify potential quality offset phenomena, clarify the influencing path and result performance, and realize real-time graded warning and trend tracing of the regulator quality status. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of the method of the present invention;

[0042] Figure 2 This is a detailed flow chart of step S1 of the present invention;

[0043] Figure 3 This is a schematic diagram of a detailed process of step S2 of the present invention;

[0044] Figure 4 This is a detailed flow chart of step S3 of the present invention;

[0045] Figure 5 This is a detailed flow chart of step S4 of the present invention;

[0046] Figure 6 This is a detailed flow chart of step S5 of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0048] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction, and therefore should not be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0049] See also Figure 1 The present invention provides a technical solution: a method for real-time monitoring of the production quality of a boiler water conditioner, comprising the following steps:

[0050] S1: Collect state data from the boiler water conditioner production process and construct a transition sequence between states. Extract the evolution characteristics of each abnormal state in the transition sequence to obtain an abnormal feature dataset.

[0051] S2: Prioritize the evolution characteristics of each type of abnormal state in the abnormal feature dataset to obtain a state priority sequence;

[0052] S3: Extract the abnormal state items with the specified priority in the state priority sequence, collect the liquid level disturbance data after the boiler water conditioner addition response during the time period when the abnormal state occurs, calculate the time delay of the corresponding liquid level disturbance recovery, and construct the liquid level disturbance delay data set;

[0053] S4: Analyze the consistency between the state priority sequence and the score change trend and the disturbance recovery time trend under the same abnormal state item in the liquid level disturbance delay dataset, determine the synchronization offset characteristics of the response period, and construct the response synchronization offset set;

[0054] S5: judging the change pattern of the regulator reaction efficiency based on the response synchronization offset set, and obtaining the boiler water regulator quality status monitoring result;

[0055] The abnormal feature dataset includes the evolution path length, state switching density, and transfer direction stability. The state priority sequence specifically includes the scoring sequence, priority label, and sorting position. The liquid level disturbance delay dataset includes the disturbance start and end time, liquid level change trend, and disturbance recovery section. The response synchronization offset set specifically includes the trend difference type, offset state label, and time period index. The boiler water conditioner quality status monitoring results include the reaction efficiency label, state change type, and abnormal response level.

[0056] See also Figure 2 , the specific steps of S1 are:

[0057] S101: Collect state data on concentration fluctuations, temperature deviations, and flow rate anomalies during the production process of boiler water conditioner, record the continuous residence time of each state on the time axis and the sequence of adjacent states, arrange the state flow in chronological order, identify state transition relationships, and obtain a continuous state transition sequence;

[0058] To collect status data on concentration fluctuations, temperature offsets and flow rate anomalies during the production process of boiler water conditioners, first, sensors are deployed at key measuring points in the production process to collect real-time data. For example, online conductivity meters, thermocouple thermometers and electromagnetic flowmeters are set at the regulator injection point, reaction tank outlet and circulation reflux section to monitor concentration, temperature and flow rate values ​​respectively. The data is recorded once per second and uploaded to the industrial data acquisition and monitoring system (SCADA). The raw data is then summarized by minute and time-series is sorted. Concentrations that continuously fluctuate beyond the baseline range of ±10% are judged as abnormal sections, where ±10% is set based on the historical stable operating condition average (such as 10.0g / L), and the abnormal judgment threshold is ±1.0g / L, that is, if the concentration is lower than 9.0g / L or higher than 11.0g / L, it is abnormal. Similarly, the temperature offset setting The benchmark is the standard process temperature of 85°C, and the deviation of ±3°C is the judgment threshold, that is, a temperature less than 82°C or greater than 88°C is abnormal. The flow rate is centered on the set benchmark of 200L / min, and the deviation range is set to ±20L / min. According to the above settings, the system traverses the time series data through the moving window method (window size 10 minutes, step size 1 minute), identifies the continuous residence segment of each type of state on the time axis, records the start time and end time, and then constructs a state label sequence based on these time periods, such as "normal-abnormal concentration-temperature deviation-normal-abnormal flow rate", sorts them in chronological order and eliminates repeated continuous states to form a concise state flow sequence, and then uses the adjacency matrix to identify the transfer relationship between states, such as "normal→abnormal concentration", "abnormal concentration→temperature deviation", and finally obtains a continuous state transfer sequence arranged in chronological order.

[0059] S102: Calculate the state transition probability distribution and the average dwell time between states in the continuous state transition sequence using a Markov chain model, identify abnormal state segments of the continuous path characteristics, and extract the switching frequency, transition direction repeatability, and dwell time combination indicators corresponding to the segments to obtain the abnormal state evolution structure feature group;

[0060] Each abnormal condition is defined based on the time series deviation behavior of key physical quantities in the boiler water conditioner production process. Specifically, the following criteria are used: Concentration abnormality is determined based on the relative deviation between the actual system concentration and the historical stable operating mean. The baseline concentration is set to the process setpoint (e.g., 10.0 g / L). If the actual measured concentration continuously exceeds the upper and lower deviation thresholds by ±10%, i.e., below 9.0 g / L or above 11.0 g / L, it is considered a concentration abnormality. Temperature deviation is determined based on the degree of deviation of the liquid temperature in the reaction or circulation section from the set standard temperature (e.g., 85°C). The deviation tolerance is set to ±3°C. If the actual temperature is below 82°C or above 88°C, it is considered a temperature deviation. Flow rate abnormality is determined based on the degree of change in the system's internal liquid flow rate relative to the standard value (e.g., 200 L / min). The allowable deviation range is set to ±20 L / min. If the detected flow rate is below 180 L / min or above 220 L / min, it is considered a flow rate abnormality.

[0061] Using a Markov chain model, we calculate the state transition probability distribution and the average dwell time between states in a continuous state transition sequence. First, we define the state set S = {s1,s2,s3,s4}, where s1 represents "normal," s2 represents "abnormal concentration," s3 represents "temperature offset," and s4 represents "abnormal flow rate." Assuming that we extract state transition paths from the constructed state sequence "normal → abnormal concentration → temperature offset → abnormal concentration → normal → abnormal flow rate → normal," we can analyze each pair to obtain the following transition events:

[0062] s1→s2: normal turns to concentration abnormality, s2→s3: concentration abnormality turns to temperature deviation, s3→s2: temperature deviation turns to concentration abnormality, s2→s1: concentration abnormality turns to normal, s1→s4: normal turns to flow velocity abnormality, s4→s1: flow velocity abnormality turns to normal.

[0063] Based on this, construct the state transition count matrix N = [n ij ], where each element n ij Indicates state s i Transfer to state s j The number of times, for example: n 12 =1, n 14 =1: Switch from normal to abnormal concentration and abnormal flow rate once each, n 23 =1, n 21 =1: From concentration anomaly to temperature deviation and normal once, n 32 =1: From temperature deviation to concentration anomaly once, n 41 =1: The flow rate changes from abnormal to normal once.

[0064] Then calculate the state transition probability matrix P=[p ij ], where p ij Indicates that the system is in state s i When the state is transferred to s j The conditional probability of:

[0065]

[0066] Among them, p ij : From state s i Transfer to state s j The probability of n ij : Status s i →s j The number of transitions, n: the total number of states in the state set, here is 4, Status i The total number of transitions to all other states.

[0067] For example, for state s1 (normal), all its transition events include: n 12 =1(→abnormal concentration), n 14 =1 (→flow rate abnormality), so the total number of transitions to this state is Then we have: This means that starting from the "normal" state, there is a 50% probability of turning to "abnormal concentration" and a 50% probability of turning to "abnormal flow rate".

[0068] Next, calculate the steady-state residence time τ of each state i , indicating state s i The average duration of each occurrence is calculated as:

[0069]

[0070] Among them, τ i : Status s i Average residence time (unit: minutes), m i : Status s i The number of times it appears in the entire time series, t ik : The length of time that the state stays continuously when it appears for the kth time.

[0071] For example, if state s2 (abnormal concentration) occurs three times and lasts for 6 minutes, 9 minutes, and 7 minutes respectively, then t 21 =6, t 22 =9, t 23 =7, then the average residence time of this state is: minutes; In summary, the state transition probability p ij reflects the directional preference of a state migration behavior, and the steady-state residence time τi It reflects the time dimension performance of each state. These two parameters together constitute the basic data of state transition dynamics in Markov process modeling.

[0072] S103: constructing a three-dimensional feature set with the state type as the index based on the abnormal state evolution structure feature group to obtain an abnormal feature data set;

[0073] According to the abnormal state evolution structure feature group, the identified abnormal state fragments are first classified and organized. Each fragment contains clear state transition start and end types, residence time length and state switching direction characteristics. Then, this information is reconstructed based on the classification of the state itself, and the state type is set as the main index. For example, "concentration anomaly", "temperature offset" and "flow rate anomaly" are used as the basis of the index dimension respectively. Then, the evolution characteristics corresponding to each state are grouped under it, including multiple dimensional data such as the transfer direction before and after the state occurs, the average residence time, the time interval between switching between the previous and next states, and the frequency of repeated switching. The data of each dimension are then stacked according to the order of state occurrence. Each group of abnormal paths forms a composite structure containing multiple attributes such as state type, switching path, and residence information. Finally, the features of all fragments are combined and summarized to form a data set with clear classification index and multi-dimensional description capabilities.

[0074] See also Figure 3 , the specific steps of S2 are:

[0075] S201: Obtain the switching frequency, transfer direction repeatability, and dwell time combination index corresponding to the abnormal state in the abnormal feature data set, perform interval normalization processing on each feature item, and obtain a standardized evolution feature set;

[0076] After obtaining the combined indicators of switching frequency, transfer direction repeatability and residence time corresponding to each abnormal state in the abnormal feature data set, we first perform a normalization operation on each feature data, and divide each feature value into several intervals according to its category. By analyzing the distribution range of similar features in the entire data set, we extract their maximum, minimum and median offset segments, and construct a unified standard interval set. Then, we map the original feature values ​​to a unified standard scale according to their intervals. For example, switching frequency can be divided into three categories: low frequency, medium frequency and high frequency according to the number of occurrences. The repeatability of transfer direction can set the repetition level interval according to the number of times the same direction appears in a short period of time. The residence time can be divided into short residence, medium residence and long residence according to the degree to which it is away from the average residence benchmark time. All feature values ​​are projected onto a unified evaluation scale in the same way to form a standardized evolution feature set.

[0077] S202: Based on the standardized evolution feature set, a weight coefficient of each combined indicator is set, and a priority score of each type of abnormal state is calculated through a weighted hierarchy function to obtain an abnormal state score result;

[0078] Based on the standardized evolution feature set, the weight coefficient of each combined indicator is set, and the priority of each type of abnormal state is evaluated by constructing a weighted hierarchical scoring function. Three standardized feature items are set: switching frequency, transfer direction repeatability, and residence time, which are represented by switching frequency F, transfer direction repeatability R, and residence time T respectively. Each feature value has been normalized to a standardized range of zero to one. In order to reflect the impact of different features on the importance of the anomaly, a weight parameter is set: switching frequency weight w F , transfer direction repeatability weight w R , residence time weight w T , must satisfy w F +w R +w T =1.

[0079] The weighting is determined as follows: a higher switching frequency indicates greater state instability and a greater cumulative risk of system disturbances, so switching frequency has the highest weight in the scoring function. A higher degree of repeatability in the transfer direction indicates a more repetitive path for the abnormal state, suggesting the presence of a mechanistic loop in the system, so it has the second highest weight. While dwell time measures the persistent impact of a single abnormal state on the process, a low switching frequency means the overall impact is relatively manageable, so it has the lowest weight. The initial recommended weight combination is: switching frequency with a weight of 0.4, transfer direction repeatability with a weight of 0.35, and dwell time with a weight of 0.25.

[0080] In order to enhance the responsiveness of the scoring model to feature volatility, the offset weight adjustment coefficient λ is introduced. This coefficient is used to dynamically adjust the static weight so that when the scoring function processes a strongly deviated feature value, it can automatically increase the influence of the item in the total score. For example, when the switching frequency of a certain state is much higher than the median of all states, it means that it is more stable, and this indicator will receive a higher weight in the scoring. Its role is reflected in the scoring formula through The form is introduced, where X represents the standardized eigenvalue, Indicates the median of the feature.

[0081] The setting basis of the offset weight adjustment coefficient λ is:

[0082] When the distribution of abnormal state characteristics in the system is relatively concentrated, that is, the characteristic standard deviation is small, which means that the overall offset of each state is not obvious, a smaller λ should be set, such as 0.2 to 0.3;

[0083] When the feature distribution is highly dispersed, that is, the standard deviation is high, it means that some states deviate from the mainstream mode. It is recommended to increase λ and set it between 0.5 and 0.7 to enhance the scoring model's ability to respond to strong deviations.

[0084] In this example, if the standard deviation of the switching frequency in the sample set is 0.25, which is a medium deviation, it is recommended to set λ=0.5.

[0085] Finally, the dynamic weighted scoring function S is constructed as follows:

[0086]

[0087] Among them, F, R, and T are the normalized values ​​of the switching frequency, transfer direction repeatability, and dwell time of the state, respectively. are the medians of switching frequency, transfer direction repeatability, and dwell time in all abnormal state samples, respectively. F ,w R ,w T are the static initial weights of switching frequency, transfer direction repeatability, and dwell time, respectively. λ is the offset weight adjustment coefficient, which is used to enhance the score sensitivity when the eigenvalue deviates. S is the priority score of the abnormal state. The larger the value, the higher the comprehensive abnormality.

[0088] Assume that the standardized characteristics of a certain abnormal state are: switching frequency F = 0.9, transfer direction repeatability R = 0.6, and dwell time T = 0.3. The median of the three characteristics in all state samples is: switching frequency median Median repeatability of transfer direction Median length of stay Use weight w F =0.4, w R =0.35, w T =0.25, assuming λ = 0.5, substitute into the formula to calculate:

[0089] S=(0.4+0.5·|0.9-0.5|)·0.9+(0.35+0.5·|0.6-0.4|)·0.6+(0.25+0.5·|0.3-0.4|)·0.3=0.90.

[0090] This abnormal state received a final priority score of 0.90, placing it at a high priority in the subsequent sorting phase. This scoring model not only reflects the fundamental importance of each feature, but also adapts to actual abnormal fluctuations. By using offset weight adjustment coefficients, it flexibly adjusts the impact of each feature on the overall score, achieving a more responsive priority assessment.

[0091] S203: Based on the abnormal state scoring results, sort each abnormal state from high to low according to the score to obtain a state priority sequence;

[0092] Collect the priority scores corresponding to all abnormal states, bind each abnormal state to its score one by one to form a score pair set, and then perform a sorting operation on the set from high to low. The sorting method can use a more stable sorting algorithm, such as general methods such as quick sort or heap sort. During the sorting process, metadata such as the original type of the state, occurrence time, and the index of the fragment to which it belongs are retained to ensure that the sorting result not only contains the score ranking but can also be traced back to the specific state entity. After the sorting is completed, the states are arranged in sequence to form a state priority sequence. The states with higher scores are at the front of the sequence, which means that the abnormal performance of the state in the system is more prominent.

[0093] See also Figure 4 , the specific steps of S3 are:

[0094] S301: Extracting the abnormal state number of the specified order in the state priority sequence, searching for the corresponding abnormality occurrence time period, and obtaining the start and end time points of the boiler water conditioner addition operation within the target time period, determining the positioning window for the addition response, and obtaining the response time interval corresponding to the abnormal state;

[0095] First, locate the time period in which the abnormal state corresponding to each number occurs. Rapid retrieval can be achieved by establishing a mapping index between the state number and the time period in the abnormal feature data set. After determining the start and end times of each abnormal period, further combine the boiler water regulator addition operation log or the historical execution record of the control system to find all regulator addition behaviors that occurred within the time period, extract the start and end time points of the operations that overlap or are adjacent to the abnormal state, and form a response addition time window for the abnormal state. For example, if an abnormal state occurs from 08:10 to 08:24, and the regulator addition record is from 08:08 to 08:12, then the response time interval of this state can be determined as 08:08 to 08:24.

[0096] S302: Based on the response time interval corresponding to the abnormal state, collect the height change sequence recorded by the liquid level sensor in the corresponding time period, calibrate the time node between the disturbance start position and the first stable platform position, and obtain disturbance recovery response section data;

[0097] Based on the response time interval corresponding to the abnormal state, the height change data continuously recorded by the liquid level sensor during this time period is retrieved. First, the starting point of the disturbance in the liquid level curve is identified through a differential algorithm, that is, the position where the liquid level value begins to rise or fall rapidly compared with the previous moment. Then, according to the trend of the liquid level value change rate gradually slowing down and tending to be stable, the first stable platform is identified, that is, the starting time point where the liquid level change rate continuously remains within a very small range. The time distance between the disturbance starting point and the stable platform is calculated as the time domain range of the disturbance recovery path. The liquid level change sequence within this time period is extracted to form the disturbance recovery response segment data. For example, if the disturbance starts at 08:10 and the stable platform is formed at 08:18, the disturbance recovery response segment is the liquid level data between 08:10 and 08:18.

[0098] S303: Calculate the time interval for the liquid level to fall from the disturbance peak to the plateau state based on the disturbance recovery response segment data, establish the corresponding disturbance recovery delay according to the state number, and obtain the liquid level disturbance delay data set;

[0099] Based on the disturbance recovery response section data, the position of the maximum liquid level deviation value is first identified as the disturbance peak, and then the starting point of the liquid level value formation stage corresponding to the recovery platform is identified. The time interval between the two is calculated, that is, the time it takes for the liquid level to fall from the disturbance peak to the stable platform. This time length is defined as the disturbance recovery delay corresponding to the abnormal state. For example, if the disturbance peak occurs at 08:12 and the platform is formed at 08:18, the disturbance recovery delay of this state is 6 minutes. All state numbers are matched and calculated in sequence, and the corresponding relationship between the abnormal state number and the corresponding liquid level disturbance recovery delay is established to form a liquid level disturbance delay dataset.

[0100] See also Figure 5 , the specific steps of S4 are:

[0101] S401: Based on the state priority sequence and the abnormal state number in the liquid level disturbance delay data set, extract the score and disturbance recovery time under the corresponding abnormal state item, and pair them in the order of the numbers to construct a state score and disturbance delay pairing group;

[0102] Based on the state priority sequence and the abnormal state numbers in the liquid level disturbance delay dataset, the numbers appearing in the two data sources are firstly intersected and matched to ensure that each number has a valid record in the scoring results and recovery delay. For each abnormal state number that meets the conditions, its corresponding score value and liquid level disturbance recovery time are extracted respectively, and a pairing structure is constructed with the number as the index. The two values ​​are bound and stored as a group of score delays, and then the whole group is sorted according to the original sorting order of the numbers, so that the formed state score and disturbance delay pairing group has a stable number reference relationship for subsequent trend comparison operations. For example, the state number 1 has a score of 0.78 and a corresponding recovery delay of 8 minutes. The state number 2 has a score of 0.85 and a recovery delay of 6 minutes. After being organized in sequence, a complete pairing group is formed.

[0103] S402: Based on the state score and disturbance delay pairing group, the score change trend and the disturbance recovery time change trend are sequentially compared to identify whether the change directions of adjacent abnormal state items in the two trend dimensions are consistent, record the state numbers with opposite trend directions or inconsistent change trends, and generate a trend consistency offset identification set;

[0104] Based on the state score and disturbance delay pairing group, the change trends of the score values ​​and recovery delay values ​​between the two items before and after are compared one by one in the order of numbers, and whether the score value increases or decreases is recorded. At the same time, it is recorded whether the recovery delay time is shortened or extended. When it is found that the score value is on an upward trend but the corresponding recovery delay is extended, or the score value decreases but the recovery delay is shortened, it is judged that the score trend corresponding to the state number is inconsistent with the delay trend direction, and the number is marked as a trend offset point. The backward iteration is continued until all adjacent pairs are compared, and all state numbers with inconsistent trend directions are collected and summarized to form a trend consistency offset identification set, which is used to mark the abnormal synergistic relationship between the score and liquid level recovery of these abnormal states. For example, the score numbered 3 increases but the delay time also increases, then its trend direction is opposite and the offset needs to be recorded.

[0105] S403: Based on the trend consistency offset identification set, mark the time period information corresponding to the abnormal state item where the offset occurs, summarize the state number, offset type and corresponding time index, and generate a response synchronization offset set;

[0106] According to the state number in the trend consistency offset identification set, trace back to the corresponding time period records in the original abnormal state sequence in sequence, extract the abnormal state time range corresponding to each offset number, including the start and end time of the anomaly, and build a time index on this basis. Bind the offset type with the state number to form a complete set of response offset annotation items. Each item includes the abnormal state number, its offset type description in the trend dimension, and the actual time segment index value. Finally, all offset items are summarized and output as a response synchronization offset set.

[0107] See also Figure 6 , the specific steps of S5 are:

[0108] S501: Call the time period index in the response synchronization offset set, match the abnormal state item with the corresponding number, extract the concentration change and flow rate change rate corresponding to the target time period in the abnormal state evolution structure feature group, and generate a state association feature comparison table;

[0109] After calling the time period index in the response synchronization offset set, match the abnormal state number in the set one by one, search for the records of the corresponding time period in the abnormal state evolution structure feature group, and extract the two core indicators of concentration change value and flow rate change rate. The calculation of the concentration change value can be based on the difference in concentration readings at the start and end times of the state and the length of time. For example, if a state number occurs between 08:10 and 08:20, and the concentration changes from 9.8g / L to 11.0g / L, the concentration change value is 1.2g / L; the flow rate change rate is calculated based on the flow rate difference and duration within the period. For example, if the flow rate increases from 195L / min to 215L / min, the change value is 20L / min, and the time is 10 minutes, then the change rate is 2.0L·min -1 The two values ​​obtained in this way constitute the reaction process characteristics of the state, and then the number, time period, concentration change value and flow rate change rate are combined and stored in the state association characteristic comparison table.

[0110] S502: Based on the state-related feature comparison table, perform combination relationship matching on each set of concentration change values ​​and flow rate change rates to determine the coordination and deviation of parameter fluctuations, classify them as reaction acceleration type, reaction lag type, or fluctuating indeterminate type, and generate a reaction efficiency change pattern label group;

[0111] Based on each set of concentration change values ​​and flow rate change rates in the state association feature comparison table, relationship matching analysis is performed one by one to identify the synergy or deviation trends between parameters, and classification processing is performed based on the judgment logic. For example, when the concentration change value is 1.2g / L and the flow rate change rate is 2.0L·min -1, both of them are rising in a positive direction and with a large amplitude, which can be judged as an accelerated reaction; if the concentration change value of the other state is 0.4g / L but the flow rate change rate is as high as 2.5L·min -1 , indicating that the system flow rate responds quickly first and the concentration change lags behind, which can be judged as a delayed response type; for example, the concentration change in a certain state is -0.6g / L (decreasing) and the flow rate change rate is 1.8L·min -1 (Increasing), the two directions are inconsistent, and the change relationship is unclear, which is classified as fluctuating and indeterminate. Finally, these judgment results are added to the original state number record to generate a set of change pattern labels with reaction efficiency classification attributes.

[0112] S503: Based on the reaction efficiency change pattern tag group, the change type, time period index, and offset characteristics of each abnormal status item are integrated to establish a state-level quality monitoring tag and generate a boiler water conditioner quality status monitoring result;

[0113] Based on the generated reaction efficiency change pattern label set, the pattern label corresponding to each abnormal state number is integrated with its time period index and trend offset characteristics to generate a complete state-level quality monitoring label. For example, for a state number with a time period of 08:10 to 08:20, an accelerated reaction efficiency type, a score that aligns with the disturbance trend, and no offset is marked, the label content is: "Number X, accelerated type, no offset, time period 08:10-08:20." Another state number with a time period of 08:22 to 08:30 is classified as a lagging type, with a decreasing score trend but increasing delay time, marking it as an offset. The label content is: "Number Y, lagging type, offset present, time period 08:22-08:30." By annotating all abnormal states one by one in this way, a structured boiler water conditioner quality state monitoring result is ultimately formed.

[0114] A real-time monitoring system for the production quality of boiler water conditioner, the system comprising:

[0115] Abnormal evolution extraction module: This module collects state data from the boiler water conditioner production process and constructs a transition sequence between states. It then extracts the evolutionary features of each abnormal state in the transition sequence to obtain an abnormal feature dataset.

[0116] State priority evaluation module: Priority scoring is performed on the evolution characteristics of each type of abnormal state in the abnormal feature dataset to obtain a state priority sequence;

[0117] Disturbance Delay Analysis Module: This module extracts abnormal status items with a specified priority from the status priority sequence, collects liquid level disturbance data after the addition of boiler water conditioner during the time period when the abnormal status occurs, calculates the time delay of the corresponding liquid level disturbance response, and constructs a liquid level disturbance delay dataset.

[0118] Response consistency judgment module: Analyzes the consistency between the state priority sequence and the score change trend and the disturbance recovery time trend under the same abnormal state item in the liquid level disturbance delay dataset, determines the synchronization offset characteristics of the response period, and constructs a response synchronization offset set;

[0119] Efficiency change identification module: determines the change pattern of the regulator reaction efficiency according to the response synchronization offset set, and obtains the boiler water regulator quality status monitoring result.

[0120] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for real-time monitoring of the production quality of a boiler water conditioner, characterized in that: The following steps are involved: S1: Collect state data from the boiler water conditioner production process and construct a transition sequence between states. Extract the evolution characteristics of each abnormal state in the transition sequence to obtain an abnormal feature dataset. S2: Prioritize the evolution characteristics of each type of abnormal state in the abnormal feature dataset to obtain a state priority sequence; S3: extracting abnormal state items of a specified priority in the state priority sequence, collecting liquid level disturbance data after the addition of boiler water conditioner during the time period when the abnormal state occurs, calculating the time delay of the corresponding liquid level disturbance recovery, and constructing a liquid level disturbance delay data set; S4: Analyze the consistency between the state priority sequence and the score change trend and the disturbance recovery time trend under the same abnormal state item of the liquid level disturbance delay data set, determine the synchronization offset characteristics of the response period, and construct a response synchronization offset set; S5: judging a change pattern of the regulator reaction efficiency according to the response synchronization offset set, and obtaining a monitoring result of the boiler water regulator quality status.

2. The method for real-time monitoring of the production quality of boiler water conditioner according to claim 1, characterized in that: The abnormal feature data set includes the evolution path length, state switching density, and transfer direction stability; the state priority sequence specifically includes the scoring sequence, priority label, and sorting rank; the liquid level disturbance delay data set includes the disturbance start and end time, liquid level change trend, and disturbance recovery section; the response synchronization offset set specifically includes the trend difference type, offset state label, and time period index; the boiler water conditioner quality status monitoring results include the reaction efficiency label, state change type, and abnormal response level.

3. The method for real-time monitoring of the production quality of boiler water conditioner according to claim 1, characterized in that: The specific steps of S1 are: S101: Collect state data on concentration fluctuations, temperature deviations, and flow rate anomalies during the production process of boiler water conditioner, record the continuous residence time of each state on the time axis and the sequence of adjacent states, arrange the state flow in chronological order, identify state transition relationships, and obtain a continuous state transition sequence; S102: Calculating the state transition probability distribution and the average dwell time between states in the continuous state transition sequence using a Markov chain model, identifying abnormal state segments of the continuous path characteristics, and extracting the switching frequency, transition direction repeatability, and dwell time combination indicators corresponding to the segments to obtain an abnormal state evolution structure feature group; S103: Constructing a three-dimensional feature set with the state type as the index based on the abnormal state evolution structure feature group to obtain an abnormal feature data set.

4. The method for real-time monitoring of the production quality of boiler water conditioner according to claim 1, characterized in that: The specific steps of S2 are: S201: Obtain the switching frequency, transfer direction repeatability, and dwell time combination index corresponding to the abnormal state in the abnormal feature data set, perform interval normalization processing on each feature item, and obtain a standardized evolution feature set; S202: Based on the standardized evolution feature set, a weight coefficient of each combination indicator is set, and a priority score of each type of abnormal state is calculated by a weighted hierarchy function to obtain an abnormal state score result; S203: According to the abnormal state scoring result, each abnormal state is sorted from high to low according to the score to obtain a state priority sequence.

5. The method for real-time monitoring of boiler water conditioner production quality according to claim 1, characterized in that: The specific steps of S3 are: S301: extracting the abnormal state number of the specified order in the state priority sequence, searching for the corresponding abnormality occurrence time period, obtaining the start and end time points of the boiler water conditioner addition operation within the target time period, determining the positioning window for the addition response, and obtaining the response time interval corresponding to the abnormal state; S302: Based on the response time interval corresponding to the abnormal state, collect the height change sequence recorded by the liquid level sensor in the corresponding time period, calibrate the time node between the disturbance start position and the first stable platform position, and obtain disturbance recovery response section data; S303: Calculate the time interval for the liquid level to fall from the disturbance peak to the plateau state based on the disturbance recovery response segment data, establish a corresponding disturbance recovery delay according to the state number, and obtain a liquid level disturbance delay data set.

6. The method for real-time monitoring of boiler water conditioner production quality according to claim 1, characterized in that: The specific steps of S4 are: S401: Based on the state priority sequence and the abnormal state number in the liquid level disturbance delay data set, extract the score and disturbance recovery time under the corresponding abnormal state item, and pair them in the order of the numbers to construct a state score and disturbance delay pairing group; S402: Based on the state score and disturbance delay pairing group, sequentially compare the score change trend with the disturbance recovery time change trend, identify whether the change directions of adjacent abnormal state items in the two trend dimensions are consistent, record the state numbers where the trend directions are opposite or inconsistent, and generate a trend consistency offset identification set; S403: According to the trend consistency offset identification set, mark the time period information corresponding to the abnormal state item where the offset occurs, summarize the state number, offset type and corresponding time index, and generate a response synchronization offset set.

7. The method for real-time monitoring of boiler water conditioner production quality according to claim 3, characterized in that: The specific steps of S5 are: S501: calling the time period index in the response synchronization offset set, matching the abnormal state item with the corresponding number, extracting the concentration change and flow rate change rate corresponding to the target time period in the abnormal state evolution structure feature group, and generating a state association feature comparison table; S502: Based on the state-related feature comparison table, perform combination relationship matching on each set of concentration change values ​​and flow rate change rates, determine the synergy and deviation degree of parameter fluctuations, classify them as reaction acceleration type, reaction lag type, or fluctuation indeterminate type, and generate a reaction efficiency change pattern label group; S503: According to the reaction efficiency change pattern tag group, the change type, time period index and offset characteristics of each abnormal state item are integrated to establish a state-level quality monitoring tag and generate a boiler water conditioner quality state monitoring result.

8. A real-time monitoring system for the production quality of boiler water conditioner, characterized in that: The method for real-time monitoring of boiler water conditioner production quality according to any one of claims 1 to 7 is implemented, wherein the system comprises: Abnormal evolution extraction module: This module collects state data from the boiler water conditioner production process and constructs a transition sequence between states. It then extracts the evolutionary features of each abnormal state in the transition sequence to obtain an abnormal feature dataset. State priority evaluation module: Priority scoring is performed on the evolution characteristics of each type of abnormal state in the abnormal feature data set to obtain a state priority sequence; Disturbance delay analysis module: extracts abnormal state items of specified priority in the state priority sequence, collects liquid level disturbance data after the addition of boiler water conditioner during the time period when the abnormal state occurs, calculates the time delay of the corresponding liquid level disturbance response, and constructs a liquid level disturbance delay data set; Response consistency judgment module: analyzes the consistency between the state priority sequence and the score change trend and the disturbance recovery time trend under the same abnormal state item in the liquid level disturbance delay dataset, determines the synchronization offset characteristics of the response period, and constructs a response synchronization offset set; Efficiency change identification module: determines the change pattern of the regulator reaction efficiency according to the response synchronization offset set, and obtains the boiler water regulator quality status monitoring result.

Images