Feces liquid dynamic proportioning regulation method based on sensing feedback

Abstract

The application discloses a manure liquid dynamic proportioning regulation and control method based on sensing feedback, and particularly relates to the technical field of waste fluid control, and comprises the following steps: a plurality of sensors are arranged in a manure liquid conveying pipeline, multi-parameter data such as flow, conductivity and solid content rate are collected, load fluctuation entropy values and control conflict count values are calculated, a current control state is determined, and the priority and execution frequency of proportional integral derivative control and adaptive control are dynamically adjusted based on the state result, a dilution ratio control signal is generated, a variable frequency pump and an electric valve are driven in linkage, clean water injection and manure liquid mixing control are realized, and the solid content rate is stabilized in a target range; the application realizes accurate regulation under steady state and disturbance by jointly identifying the system state through the entropy value and the conflict count and dynamically switching the control strategy priority and frequency; meanwhile, the dilution ratio control signal is linked with the variable frequency pump and the electric valve, a closed-loop dilution control path is constructed, and the intelligence and stability of the solid content rate regulation and control are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of waste fluid control, more particularly, to a manure liquid dynamic proportioning regulation method based on sensing feedback. BACKGROUND

[0002] With the rapid development of large-scale livestock and poultry breeding industry, the efficient treatment and resource utilization of manure liquid have become increasingly serious. In the process of transportation, deployment and subsequent treatment of high-concentration and high-variability manure liquid, accurate control of solid content has become a key factor affecting the efficiency of biogas fermentation, irrigation safety and land carrying capacity. How to achieve automatic proportioning and stable dilution of manure liquid in a dynamically disturbed environment is a technical bottleneck that needs to be broken through in the field of agricultural ecological governance and intelligent agricultural equipment.

[0003] Most of the existing manure liquid regulation systems use fixed proportion of water injection or manual intervention type liquid proportioning method, which usually relies on experience to set the dilution ratio and controls the pump valve opening and closing at a fixed pace, lacking the ability to finely perceive and respond to the real-time state of the slurry. In the process of manure liquid transportation, the key parameters such as slurry concentration, conductivity, flow rate, etc. are often subjected to severe disturbance due to changes in terrain, fluctuations in pump stations, upstream injection rhythm or pipe deposition conditions. The traditional fixed logic control method cannot timely adapt to changes in system state, which easily leads to the expansion of solid content deviation, causing reduced fermentation efficiency, aggravated pipe blockage, and even crop root burn, land eutrophication and other chain risks.

[0004] To improve the automatic control capability, some studies attempt to introduce proportional-integral-derivative control or fuzzy logic algorithm to realize closed-loop adjustment of dilution ratio. However, due to the multiple disturbance dimensions, strong execution lag and complex control response path of the manure liquid system, the above methods are prone to control conflicts and logic oscillation problems in the scenario of multiple control logic superposition. For example, in the context of rapid fluctuations in slurry concentration and sudden increase in flow rate in the mixing zone, the proportional-integral-derivative controller and the adaptive adjustment algorithm may repeatedly switch the dominant right, causing adjustment frequency resonance phenomenon, and then causing frequent opening and closing of the pump valve, flow shock or pipe pressure mutation, which seriously weakens the running stability and control accuracy of the system. Therefore, the present application proposes a manure liquid dynamic proportioning regulation method based on sensing feedback to solve the above problems. SUMMARY

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] The manure liquid dynamic proportioning regulation method based on sensing feedback comprises the following steps:

[0007] A plurality of sensors are arranged in the manure liquid transportation pipeline to collect flow rate, conductivity, solid content, pressure and temperature data;

[0008] Adopting multi-dimensional parameter joint analysis method, the load fluctuation entropy value in the set time window is calculated, which is used to represent the disturbance intensity and complexity of system operation.

[0009] The proportional integral differential control output and the adaptive control output are recorded synchronously, and the conflict frequency between the two types of control instructions is counted in the time window to generate the control conflict count value.

[0010] According to the combination of the load fluctuation entropy value and the control conflict count value, the current control state is determined, which is divided into four states: high disturbance and high conflict, high disturbance and low conflict, low disturbance and high conflict, and low disturbance and low conflict.

[0011] According to the control state type, the priority and control execution frequency of the control strategy are dynamically adjusted. In the high disturbance and high conflict state, the single proportional integral differential control is adopted and the frequency is reduced. In the low disturbance and low conflict state, the proportional integral differential control and the adaptive control are allowed to run cooperatively and the frequency is increased. In the remaining states, the switching is performed according to the preset rules.

[0012] The dilution ratio control signal is outputted to link the frequency conversion pump and the electric valve to execute the clean water injection and the fecal slurry mixing, and to execute the mixing control within the target solid content rate range.

[0013] In a preferred embodiment, when the sensor is used, a time-sharing synchronous acquisition mechanism is adopted to trigger the data reading of all sensors according to a unified time reference, obtain a complete parameter set at the corresponding time point, and pre-set a data validity determination rule to eliminate abnormal values and missing values generated in the acquisition, forming a continuous parameter sequence that can be used for subsequent load fluctuation entropy value calculation.

[0014] In a preferred embodiment, the following processing steps are included before calculating the load fluctuation entropy value:

[0015] The acquired flow rate, conductivity, solid content rate, pressure and temperature data are normalized to form a multi-dimensional parameter input vector. The multi-dimensional parameter input vector is segmented according to the set time window to construct a multi-dimensional time sequence parameter matrix. The parameter matrix is filtered for noise and abnormal values to generate an effective data set for entropy value calculation.

[0016] In a preferred embodiment, the calculation of the load fluctuation entropy value includes the following steps:

[0017] A joint probability distribution model is established for the effective data set using the kernel density estimation method, and then the Shannon entropy value is calculated based on the joint probability model to obtain a unique load fluctuation entropy value. The refresh period and window update strategy of the entropy value output are set to update the load fluctuation entropy value in real time.

[0018] In a preferred embodiment, the generation of the control conflict count value includes the following steps:

[0019] The control instruction sequence outputted by the proportional-integral-derivative control strategy and the adaptive control strategy respectively is recorded within a set time window, a control conflict determination criterion is set, and the directions and amplitudes of the two types of control instructions at the same time point are compared. When the adjustment directions are opposite and the amplitude difference exceeds a first conflict threshold, it is counted as a complete control conflict event. When the adjustment directions are consistent but the amplitude difference exceeds a second conflict threshold, it is converted into a weighted conflict number according to the difference. All complete control conflict events and weighted conflict numbers are accumulated within the time window to form a control conflict count value, and the value is dynamically updated with the time window sliding.

[0020] In a preferred embodiment, the time window in the process of generating the load fluctuation entropy value and the control conflict count value is consistent.

[0021] In a preferred embodiment, determining the current control state includes the following steps:

[0022] A disturbance threshold interval of the load fluctuation entropy value and a conflict threshold interval of the control conflict count value are set, and the entropy value and the count value are mapped into a disturbance level and a conflict level, respectively.

[0023] A two-dimensional state mapping matrix is constructed, and the disturbance level and the conflict level are combined and corresponded to four types of control states, wherein the disturbance level is divided into high disturbance and low disturbance, and the conflict level is divided into high conflict and low conflict.

[0024] A state retention mechanism is used to judge the consistency of the state results of consecutive multiple time windows. If the state remains unchanged for more than a set period, the state change is confirmed to be valid.

[0025] In a preferred embodiment, when the determined control state does not last for a set confirmation period, a historical state record window containing the current time point is extracted, and the occurrence frequencies of four states, i.e., high disturbance high conflict, high disturbance low conflict, low disturbance high conflict, and low disturbance low conflict, in the window are counted.

[0026] The probability value of the current state to be confirmed in the historical window is calculated and compared with a preset probability threshold.

[0027] When the historical probability of the current state to be confirmed is greater than or equal to the preset probability threshold, the current state to be confirmed is directly confirmed as the current control state, otherwise, the last confirmed state remains unchanged.

[0028] In a preferred embodiment, dynamically adjusting the priority and control execution frequency of the control strategy includes the following steps:

[0029] The mapping relationship between the control state and the execution cycle is established, the high disturbance and high conflict state is corresponded to the lowest control frequency interval, the low disturbance and low conflict state is corresponded to the highest control frequency interval, and the specific control cycle value corresponding to each type of state is set;

[0030] When the control state is high disturbance and high conflict, the adaptive control strategy execution is suspended, the proportional integral derivative control cycle is extended to more than twice the basic cycle, when the control state is low disturbance and low conflict, the parallel calculation module of proportional integral derivative control and adaptive control is activated, and the control cycle is compressed to within 50% of the basic cycle, and the remaining state selects the intermediate control cycle according to the mapping rule and switches according to the set strategy.

[0031] The technical effects and advantages of the present application are:

[0032] Based on the joint analysis of load fluctuation entropy value and control conflict count value, the present application establishes a dual recognition mechanism of disturbance intensity and control consistency, can divide the complex dynamic behavior that may appear in the manure liquid regulation process into four typical states, and then drives the dynamic switching of the control strategy. The design not only breaks the limitation of fixed frequency output of the traditional proportional integral derivative control strategy, but also avoids the regulation instability problem caused by adaptive control in high conflict environment. By distinguishing different operating states and setting response rules, the system can adaptively adjust the priority and frequency of the control strategy under different disturbance and conflict conditions, so that the control system not only has anti-interference ability, but also can maintain rapid response, improves the stability and flexibility of regulation.

[0033] The present application uses the dilution ratio control signal as a bridge to accurately convert multi-source sensing information and control strategy output into coordinated driving instructions of variable frequency pump and electric valve, so that the clean water injection and manure liquid mixing process no longer depends on fixed proportion or artificial setting parameters, but is automatically executed by the system according to the current state dynamic calculation control. This method not only realizes the closed-loop regulation of slurry solid content, but also can maintain stable output in scenes with severe disturbance or rapid change of solid content, effectively inhibiting energy waste or concentration deviation caused by excessive dilution or control delay. Through the dynamic generation of dilution signal and the linkage of physical actuator, the system has higher intelligent level and on-site adaptability, and is suitable for complex working condition environment with sensitive fluctuation of solid content requirements. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to facilitate the understanding of those skilled in the art, the present application will be further described below with reference to the drawings;

[0035] Figure 1 The principle diagram of the manure liquid dynamic proportioning regulation method based on sensing feedback in the present application. DETAILED DESCRIPTION

[0036] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0037] With reference to Figure 1 The following examples are obtained:

[0038] Embodiment 1: A manure liquid dynamic proportioning control method based on sensing feedback, comprising the following steps:

[0039] A plurality of sensors are arranged in the manure liquid conveying pipeline to collect flow rate, conductivity, solid content, pressure and temperature data; the physical and chemical parameter states of the slurry in the conveying process are comprehensively sensed to form a multi-dimensional and high-resolution data basis, which provides input support for subsequent disturbance sensing, control judgment and dilution ratio calculation.

[0040] A multi-dimensional parameter joint analysis method is used to calculate the load fluctuation entropy value in a set time window, which is used to represent the disturbance intensity and complexity of system operation; through statistical modeling and information entropy calculation on the above-mentioned multiple parameters, the intensity of slurry state change and the dynamic uncertainty of system operation are quantified, thereby providing a judgment basis for sensitivity adjustment and disturbance identification of the control strategy.

[0041] The proportional-integral-derivative control output and the adaptive control output are recorded synchronously, the conflict frequency between the two types of control instructions is counted in the time window, and the control conflict count value is generated; the inconsistency generated in the actual execution between different control strategies is quantified, and whether there is a coordination problem or a priority conflict in the current control chain is identified, so as to support the optimization switching of the subsequent control logic.

[0042] According to the combination of the load fluctuation entropy value and the control conflict count value, the current control state is determined, and divided into four states of high disturbance and high conflict, high disturbance and low conflict, low disturbance and high conflict, and low disturbance and low conflict; through a two-dimensional state mapping mechanism, the disturbance amplitude and control consistency of the system are jointly quantified, a four-quadrant control state space is constructed, a state basis is provided for subsequent control priority setting and frequency adjustment, and a control scheduling mechanism with better situational awareness is realized.

[0043] According to the control state type, the priority and control execution frequency of the control strategy are dynamically adjusted. In a high disturbance and high conflict state, a single proportional integral derivative control is adopted and the frequency is reduced. In a low disturbance and low conflict state, the proportional integral derivative control is allowed to run cooperatively with adaptive control and the frequency is increased. In other states, switching is performed according to a preset rule. The control system is adapted to different operating scenarios. In a state of strong disturbance or large instruction conflict, the control complexity is limited and the periodic stable operation is prolonged. In a state of stable state or high control consistency, the control response speed and accuracy are improved, so as to balance the system performance between stable state control and fast response.

[0044] An output dilution ratio control signal is outputted, a variable frequency pump and an electric valve are linked to execute clean water injection and manure liquid mixing, and mixing control in a target solid content range is executed. The control instruction generated after state determination and control strategy optimization is converted into an executable physical adjustment action, accurate dilution water addition and dynamic stable control of manure liquid concentration are realized, and the operation requirements of downstream applications such as anaerobic fermentation or irrigation systems are ensured.

[0045] In the embodiment, to realize high-precision, low-delay and multi-dimensional real-time monitoring of the slurry state in the manure liquid conveying process, multiple types of sensors need to be deployed at key positions of the manure liquid conveying pipeline, and time-synchronized data collection and effectiveness verification are performed on the sensor data to form high-quality and continuous multi-dimensional parameter sequences as the basis for subsequent disturbance entropy calculation and control decision-making. Representative positions on the conveying pipeline are selected to install sensors including but not limited to flow sensors, conductivity sensors, solid content sensors, pressure sensors and temperature sensors. To cover the flow field characteristics of manure liquid, flow and pressure sensors should be arranged near the pump outlet, solid content and conductivity sensors should be arranged in low-lying sections of the pipeline, and temperature sensors should be arranged in temperature-volatile areas such as buried sections or exposed sections. After the layout is completed, to ensure that the data collected by various sensors have time alignment capability, a unified time reference sampling framework is constructed, a global sampling frequency (for example, every 5 seconds is a sampling period) is set, at the start of each sampling period, a data collection trigger signal is uniformly sent, and various sensors logically "simultaneously" complete data reading operations to form a complete multi-parameter data group corresponding to the same time point.

[0046] To achieve the "time-synchronized" acquisition, during the data acquisition execution, the acquisition time window is set for each type of sensor, for example, a maximum of 0.5 seconds of sampling response delay is allowed for each sampling period, and a time stamp is bound for each type of sensor in the data buffer, ensuring that all sensor data falls within the current sampling window time period. If a sensor fails to return data within the current window or the delay exceeds the set upper limit, it is marked as data missing. For successfully acquired data, it is stored as a five-dimensional structure data set, including flow value (such as 62.3 L / min), conductivity value (such as 12.5 mS / cm), solid content value (such as 7.8%), pressure value (such as 1.35 bar) and temperature value (such as 33.7℃).

[0047] After forming the basic parameter data set, to further ensure the availability and stability of statistical analysis of the data, the effectiveness of the collected data set needs to be determined. This operation includes two types of processing: one is the identification of abnormal values, using the 3σ rule based on historical distribution, the historical mean and standard deviation of each parameter are dynamically calculated, if the current data point deviates from the average value beyond the set threshold (such as ±3 times the standard deviation), it is determined as an abnormal value; the second is the missing value judgment, the sensor value without returned data is directly marked as missing. For parameter groups with abnormalities or missing values, logical rule processing is used, for example, if only one data is abnormal, the median value of the previous and subsequent periods can be used for interpolation; if more than two consecutive parameters are missing, the entire data is discarded and not used for subsequent calculation.

[0048] After completing the data synchronization and effectiveness determination, the cleaned data set is sequentially spliced to form a continuous multi-parameter time sequence, which is used for subsequent disturbance analysis and control judgment. This data sequence is updated in a time window manner, for example, a window containing 10 groups of data is set, covering a time span of 50 seconds, after each new data is obtained, the window moves one position to the right, forming a high coverage and high time resolution data input stream. This continuous parameter sequence will be the only input source for subsequent load fluctuation entropy calculation, its integrity, synchronization and effectiveness directly determine the accuracy and stability of disturbance judgment and control strategy execution.

[0049] To improve the accuracy of disturbance identification in the dynamic proportioning control method of manure liquid based on sensor feedback, a series of structured processing is needed on the collected manure liquid parameter data before the load fluctuation entropy value is calculated. The processing flow includes normalization, multi-dimensional segmentation, noise removal and abnormal value cleaning, to ensure that the data input into the entropy value calculation has the characteristics of dimensionless, complete structure and real fluctuation. The purpose of unified normalization processing on the collected multiple parameter data is to eliminate the scale difference between each physical quantity, facilitating subsequent modeling calculation. For example, the unit of flow is liters per minute, the unit of conductivity is millisiemens per centimeter, the unit of solid content is percentage, the unit of pressure is bar, and the unit of temperature is degree Celsius. If they are directly used for joint analysis, it will lead to bias. In this embodiment, the 0-1 normalization formula X'=(X-Xmin) / (Xmax-Xmin) is used, where Xmin and Xmax are the minimum and maximum values of each parameter in the last 200 groups of sampling data. For example, if the flow range of a certain period is 40+80 L / min, the normalized value is (62-40) / (80-40)=0.55. Through this processing, a multi-dimensional parameter input vector with consistent structure and unified scale is formed as the basis for further modeling.

[0050] To capture the dynamic change trend of the parameters, the normalized parameter vector is segmented by time according to the sliding window, and a multi-dimensional time sequence parameter matrix with time dimension is constructed. The specific method is to set the time window length to 10 groups of data and the step length to 1, that is, the window slides forward once every time a new group of data is received, forming a new matrix segment. For example, when the time window length is set to 10 groups of data, the first time period can include data numbers 1 to 10, the second time period includes data numbers 2 to 11, and so on. Each time the window slides by one group of data, a continuous and overlapping time segment is formed. After such processing, a 5-dimensional parameter x 10-group time sequence two-dimensional matrix is formed in each time period, which not only maintains the instantaneous structure between parameters, but also introduces the horizontal time evolution characteristics, providing structured input for subsequent probability modeling and entropy value estimation.

[0051] After constructing the time sequence parameter matrix, noise filtering operation needs to be performed on the data in the matrix to avoid introducing false fluctuation signals due to accidental electromagnetic interference, sensor jump or signal jitter. Noise filtering is performed by combining multi-order moving average and median filtering, for example, for a certain dimension parameter in each time period, 3-point moving average and median are calculated to correct the original value. Taking a 10-segment temperature data [32.1, 32.2, 39.8, 32.0, 31.9] as an example, the 39.8℃ is obviously a short-time interference peak, which can be replaced by 32.1℃ using median filtering. This processing ensures the continuity and authenticity of the parameter change trend and eliminates the interference of short-time signal burrs.

[0052] To eliminate extreme outliers that may affect statistical calculations, distribution boundary-based anomaly detection is performed on the matrix of each time period. Specifically, within each time window, the average value and standard deviation of each dimension parameter are calculated, and if a value deviates from the average value by more than three standard deviations, it is considered an outlier. Taking conductivity as an example, if the average value is 12.5 mS / cm and the standard deviation is 0.4 mS / cm in a time period, the threshold upper and lower limits are [11.3, 13.7], and data exceeding this range will be removed or interpolated. After the above four steps of normalization, sliding segmentation, noise filtering and outlier removal, the final effective data set is formed, which is stable in structure, true in fluctuation and has no obvious distortion or mutation. It is the only input source for load fluctuation entropy value calculation, directly affecting the sensitivity of entropy value and the accuracy of control decision.

[0053] The core of the mechanism for constructing a unified time reference sampling framework, uniformly issuing data collection trigger signals, and logically simultaneously completing data reading operations by various sensors is not to require all types of sensors to read synchronously without delay at the physical level, but to achieve logical synchronous collection within a unified sampling window through timestamp calibration mechanisms, sampling scheduling design, and result buffering strategies. This ensures that all data can be classified into the same time slice when calculating, meeting the timing consistency requirements in dynamic proportioning calculation. In one implementation of the present application, to ensure the synchronicity of multi-sensor data under a unified time reference, existing unified time trigger mechanisms and timestamp buffering scheduling strategies can be used. Specifically, the main control unit can issue a sampling trigger signal, and various sensors can complete sampling operations after receiving the signal. In combination with sampling tolerance and response determination strategies, effective data can be synchronized and collected within 0.5 seconds. The above method is a mature technical means that has been widely used in the industry.

[0054] To realize quantitative measurement of the disturbance degree of the manure liquid flow state, the information entropy theory is used to construct a disturbance index based on sensor feedback. Specifically, a joint probability distribution model is established through kernel density estimation, and the Shannon entropy value is calculated based on this to obtain a unique load fluctuation entropy value that can reflect the complexity of system disturbance in real time. In practical applications, this method can adapt to non-normal, nonlinear, and multi-dimensional coupled slurry parameter behavior characteristics, and has high universality and adaptability. The joint probability model is established using the kernel density estimation method. The data set used is derived from the parameter matrix in the sliding time window after removing outliers and noise in the previous step. This matrix includes normalized flow, conductivity, solid content, pressure, and temperature data at multiple time points. To ensure the accuracy and continuity of the estimation, a Gaussian kernel function is used for density estimation, for example, for each data point , the probability density estimation is:

[0055] ; where K is the Gaussian kernel function, h is the bandwidth, d is the dimension of the parameter (5 in this embodiment), and n is the number of samples. In practical applications, the bandwidth h is automatically selected by the Silverman rule or cross-validation to avoid introducing estimation bias by manual setting. After obtaining the joint probability density model, the disturbance level in the current time window is calculated by the information amount using the Shannon entropy definition. Since the probability density function is a high-dimensional continuous function, numerical integration methods are used to approximate the integral. In this embodiment, the Monte Carlo sampling combined with grid approximation method is selected, and the equal density points are selected in the estimation space and the probability value under the distribution is calculated, and finally the approximate entropy value is obtained. For example, in a time window, the five-dimensional joint distribution constructed by kernel density estimation finally obtains a load fluctuation entropy value of 2.63 (unit: bit), which is used to represent the disturbance intensity of the current slurry state.

[0056] To realize the real-time response capability of the load fluctuation entropy value, the Shannon entropy value calculation is taken as part of the sliding time window update mechanism. After a new set of sensor data is collected, the window slides forward by one step (for example, 5 seconds), recombines the data subset, and re-executes the probability estimation and entropy value calculation process. This sliding mechanism ensures that the entropy value at each moment is evaluated based on the latest state information, realizing high time resolution disturbance monitoring. Taking continuous operation as an example, if each 10 groups of data form a window, the system will continuously output about 10 groups of entropy values within 60 seconds, which helps to accurately depict the occurrence, change and dissipation process of the disturbance. To avoid the output of entropy values being too frequent and causing control response jitter, the entropy value refresh period and window reconstruction rule are set to build a stable entropy value output strategy. The refresh period can be set according to the scene, for example, output the entropy value once every three window periods, or set a threshold mechanism, only when the entropy value changes by more than a set threshold (such as ±0.2 bit) will a new output be triggered. In this way, on the one hand, the sensitivity of disturbance identification is maintained, and on the other hand, the control chain stability is improved to prevent control misactions caused by high-frequency fluctuations.

[0057] To effectively identify the possible conflict between the proportional-integral-derivative control strategy and the adaptive control strategy in parallel operation, and improve the stability and coordination of the control chain, a conflict identification mechanism based on direction consistency and amplitude difference judgment is adopted to generate a dynamically updated control conflict count value. This count value is used for subsequent control state determination and, together with the load fluctuation entropy value, constitutes the core judgment basis in the dynamic proportioning regulation of slurry. Within a set time window, the output control instruction sequences of the proportional-integral-derivative control strategy and the adaptive control strategy are recorded respectively. To ensure the time consistency and statistical effectiveness of the data, the two types of control instructions must be recorded based on the same data sampling rhythm and refresh frequency. For example, if a control instruction is generated every 5 seconds within a certain time period, and the length of the time window is set to 60 seconds, then 12 pairs of control instruction data will be generated within each time window. Each set of data contains two values representing the dilution ratio adjustment amplitudes output by the two control strategies at the current time, for example, +3.5% for the proportional-integral-derivative strategy and +1.0% for the adaptive strategy.

[0058] The judgment criteria for control conflict are set, and the direction consistency and amplitude difference of the two types of control instructions at the same time point are judged. Direction consistency means that the adjustment targets of the two strategies at the current time are the same, both increasing or decreasing the dilution ratio; opposite directions indicate that one is increasing and the other is decreasing. The amplitude difference is defined as the absolute value of the difference between the two values. Two conflict thresholds are set: the first conflict threshold is used to judge the serious conflict when the directions are opposite, for example, set to 3.0%; the second conflict threshold is used to judge the soft conflict when the directions are consistent but the deviation is too large, for example, set to 2.0%. Taking a certain time point as an example, the proportional-integral-derivative strategy outputs +4.2%, and the adaptive strategy outputs -0.8%. The directions are opposite and the difference is 5.0%, which is greater than the first conflict threshold, so it is counted as a complete control conflict event.

[0059] When the adjustment directions are consistent but the amplitude difference exceeds the second conflict threshold, it is not directly counted as a complete conflict event, but is proportionally converted into a weighted conflict number. The conversion rule is based on the proportional relationship between the difference and the second conflict threshold for linear mapping, for example, the proportional-integral-derivative strategy outputs +6.0% and the adaptive control strategy outputs +3.0%. The difference is 3.0%, which exceeds the second threshold by 1.0%. It can be proportionally converted to (3.0-2.0) / 3.0=0.33, i.e., 0.33 times of weighted conflict. This method retains the weight influence of slight control inconsistency, while avoiding excessive amplification of non-essential differences.

[0060] All complete control conflict events in a time window are cumulatively summed with the weighted conflict times to form a control conflict count value corresponding to the window. For example, in a certain 60-second window, 3 complete conflicts with opposite directions and greater than the first threshold value occur, and 2 weighted conflicts with consistent directions but greater than the second threshold value occur, which are respectively converted to 0.4 times and 0.6 times, and then the control conflict count value in the window is 3+0.4+0.6=4.0. The count value is used as a quantitative index to represent the degree of incoordination between the current control chains. With data updates, the window slides by steps, and each time the earliest set of data is removed and the latest set of data is added, so as to realize dynamic updating of the control conflict count value, and keep the same updating pace as the load fluctuation entropy value on the time axis, thereby providing real-time input for subsequent control state mapping. The embodiment distinguishes between hard conflicts and soft conflicts by jointly judging the control instructions in the direction and amplitude dimensions, introduces a weighted conversion mechanism, so that the control conflict count value has fine expression ability and dynamic response ability, and is suitable for complex scenes of collaborative operation of multiple control strategies, especially in agricultural and environmental protection occasions with large disturbance of manure liquid flow and high regulation response requirements.

[0061] To realize joint judgment of the disturbance degree and the control chain coordination degree in the manure liquid conveying process and improve the situational awareness of the regulation process, a dual-index state evaluation method of the load fluctuation entropy value and the control conflict count value is introduced, and through the steps of grade division, two-dimensional combination and time sequence judgment, the disturbance grade and the conflict grade are combined to form four control states, i.e. high disturbance high conflict, high disturbance low conflict, low disturbance high conflict and low disturbance low conflict, and a four-class control state space is constructed as the basis for dynamic control strategy switching. According to the actual change interval of the load fluctuation entropy value and the control conflict count value, the disturbance threshold interval and the conflict threshold interval are set, and the continuous values are mapped into discrete grades respectively to form classification marks for easy determination. Taking an actual application as an example, if the historical change range of the load fluctuation entropy value is between 1.2 and 3.6 bits, the disturbance grade division threshold can be set to 2.4 bits, and the entropy value greater than or equal to 2.4 is determined as high disturbance, and less than 2.4 is determined as low disturbance. Similarly, if the control conflict count value usually fluctuates between 0 and 6 times, the conflict grade division threshold can be set to 3.0 times, and greater than or equal to 3.0 is recorded as high conflict, and less than 3.0 is recorded as low conflict. This mapping method converts complex continuous variables into control state grades, making the subsequent state combination more operable and clear in judgment.

[0062] On the basis of disturbance level and conflict level, a two-dimensional state mapping matrix is constructed. The matrix takes disturbance level as the vertical axis and conflict level as the horizontal axis. The combination of the two levels is mapped into four control states, which are high disturbance high conflict, high disturbance low conflict, low disturbance high conflict and low disturbance low conflict. For example, the load fluctuation entropy value in the current time window is 2.9 bits, and the control conflict count value is 3.8 times. Then the corresponding disturbance level is high disturbance, and the conflict level is high conflict. The corresponding control state in the mapping matrix is high disturbance high conflict. This state represents that there is significant parameter disturbance and high conflict between control chains at the current time, and a more conservative control strategy should be adopted. The four-quadrant state division method has clear logical distinguishability, which is convenient for the control strategy to switch according to the state.

[0063] After obtaining the control state of each time window, a state retention mechanism is set to judge the consistency of the determination results in continuous multiple time windows, so as to filter the short-time misjudgment caused by accidental disturbance. A state confirmation period threshold is set, for example, the state remains consistent in continuous 3 windows, and it is determined as a real state switching. Taking a 5-second sliding window as an example, the state confirmation period is 15 seconds. When a certain state appears for the first time, it does not immediately change the current control strategy, but observes whether it remains consistent in the next two time windows. If the state of the continuous three time windows is "high disturbance low conflict", the current control state is updated, otherwise the original state is maintained. This mechanism can effectively reduce the risk of frequent switching caused by short-time burst, and improve the stability of the control chain.

[0064] To solve the problem of response delay caused by frequent disturbance changes and overly conservative control strategy switching in the dynamic proportioning control process of fecal liquid, an early confirmation mechanism based on historical state probability judgment is introduced on the basis of the conventional state retention mechanism. When the current determined control state has not continuously met the confirmation period condition, the historical state sequence within a certain time span is combined for frequency statistics and probability evaluation, and the state change is confirmed in advance when the set threshold condition is met, thereby improving the responsiveness of the control strategy while maintaining system stability. A new control state determination result is generated within a certain time window, but the state has not been consistent in consecutive multiple time windows, so it has not met the confirmation period condition. In this case, the historical state record window of a fixed length containing the current time point is extracted forward, for example, all state determination results in the past 10 time windows (corresponding to 50 seconds) are selected to form an ordered state sequence dataset. Each element records a control state, such as high disturbance high conflict, high disturbance low conflict, low disturbance high conflict, or low disturbance low conflict, ensuring consistency with the control state definition system and including the latest state result of the current "to-be-confirmed state". The historical state sequence dataset is subjected to frequency statistics, and the number of occurrences of each of the four control states in the window is counted one by one. For example, in 10 time windows, the high disturbance high conflict state appears 3 times, the high disturbance low conflict state appears 2 times, the low disturbance high conflict state appears 1 time, and the low disturbance low conflict state appears 4 times. The current to-be-confirmed state is "high disturbance high conflict" state, and since it appears 3 times in the window, its historical probability can be calculated as 3 / 10, i.e. 30%. This process converts the original state label data into structured indicators that can be used for probability comparison, laying the foundation for the next step of judgment.

[0065] A historical probability threshold is set as the trigger condition for early confirmation, for example, the threshold is set to 0.6, i.e. 60%. The probability value of the current to-be-confirmed state in the historical window is compared with the threshold to determine whether it has enough trend support to be confirmed as the current effective control state in advance. If the current to-be-confirmed state appears in the history with a proportion greater than or equal to the set threshold, it means that the system has repeatedly shown this state feature over a period of time, with a strong stable trend, so it can skip the remaining confirmation period and directly update the state.

[0066] The processing logic determines the current control state according to the comparison result. When the historical probability meets the threshold requirement, the current to-be-confirmed state is directly confirmed as the new control state, and the corresponding control strategy adjustment is triggered immediately; otherwise, if the threshold is not met, the last confirmed state is maintained unchanged, and the current state continues to be observed whether it remains consistent in the subsequent time window to meet the regular confirmation period condition. To illustrate with a specific example, if the current state is "high disturbance and low conflict", it appears 6 times in 10 windows, with a probability of 60%, which meets the threshold requirement, so the state can be confirmed in advance; if it only appears 3 times, the probability is 30%, so the original state is maintained and continues to be monitored. By introducing the early confirmation mechanism based on the historical state probability, the embodiment ensures the stability of the control while enhancing the system's perception ability of the state trend, improving the response speed in the initial stage of disturbance change, avoiding missing the best regulation opportunity due to waiting for the confirmation period, and being suitable for complex application scenarios such as manure liquid flow rate, conductivity and solid content rate fluctuation or frequent transient events. The method can flexibly combine the sliding window mechanism and the judgment rule, and has good engineering practicability and scalability.

[0067] To cope with the problems of control error accumulation and regulation instability caused by the dynamic changes of parameters such as flow rate, conductivity and solid content rate during the manure liquid conveying process, a dynamic control frequency scheduling mechanism based on the control state recognition result is designed. This mechanism matches the control state with the priority of the control strategy and the execution frequency of the control instruction, realizes flexible control and intelligent switching of the regulation rhythm, and improves the response ability and resource scheduling efficiency of the system on the basis of ensuring its steady-state operation. A mapping relationship between the control state and the control execution period is established, and the four types of control states obtained by the above judgment are matched one by one with the execution period interval. Specifically, the high disturbance and high conflict state corresponds to the lowest control frequency interval (i.e. the maximum execution period), for example, the execution period is set to 20 seconds; the low disturbance and low conflict state corresponds to the highest control frequency interval (i.e. the minimum execution period), for example, the execution period is set to 5 seconds; the remaining two states-high disturbance and low conflict, low disturbance and high conflict-are set to intermediate frequency intervals, for example, 10 seconds and 15 seconds, and the mapping relationship remains stable. This mapping relationship is coded into the control logic in the form of a data table, realizing the switching of the execution frequency driven by the state change and providing a structural basis for subsequent fine control.

[0068] When the current control state is identified as a high disturbance and high conflict state, the control conservative mode is entered, the adaptive control strategy is suspended, only the proportional integral derivative control strategy is reserved for control decision, and the control command execution period is significantly extended to avoid oscillation or actuator fatigue caused by frequent adjustment. For example, under the condition that the original basic control period is 10 seconds, it is extended to 20 seconds or more, ensuring that the system can still maintain stable output when the parameters fluctuate sharply and the control strategies are inconsistent, preventing the control chain from over-reaction, causing dilution ratio overshoot or solid content fluctuation.

[0069] When the control state is identified as a low disturbance and low conflict state, the control synergistic enhancement mode is entered, the proportional integral derivative control strategy and the adaptive control strategy are activated in parallel, and the two can work together to control the output after consistency detection to improve response accuracy and self-learning ability. In this state, the control period is compressed to within 50% of the basic period, for example, from 10 seconds to 5 seconds, so that the adjustment is more timely and can finely track the slow drift trend of flow, conductivity or solid content, while ensuring adjustment accuracy, reducing resource waste and regulation delay.

[0070] For the intermediate state of high disturbance and low conflict or low disturbance and high conflict, the intermediate control period is selected according to the mapping rule described above, and the participation weight of the control strategy is set according to the current conflict degree. For example, in the low disturbance and high conflict state, the control period is set to 15 seconds, the proportional integral derivative control is dominant, and the adaptive strategy remains cold standby; in the high disturbance and low conflict state, the period can be set to 10 seconds, and the adaptive strategy is allowed to participate with a lower weight to identify trend changes in advance and make slight corrections. All strategy switching is automatically completed based on state-driven logic to avoid human intervention and improve the intelligence and stability of the control system. The present embodiment maps the disturbance degree and control consistency into the regulation rhythm and strategy priority, and constructs a state-aware control execution rhythm management method, which is suitable for typical application environments such as flow field dramatic change, concentration fluctuation, and complex control chain hierarchy in dynamic deployment of manure liquid, significantly improving the overall control accuracy, response speed and system stability.

[0071] In the present embodiment, in order to achieve the target solid content rate control in the mixing process of fecal sewage liquid, the control strategy output needs to be finally converted into a dilution ratio control signal that can drive physical execution actions, and accordingly the variable frequency pump and electric valve are linked to complete the injection of clean water and the mixing of fecal sewage liquid. The process is based on the control state determination result to dynamically determine the control strategy dominance and execution frequency, and the proportional integral derivative control strategy and the adaptive control strategy generate adjustment instructions in the cycle respectively, and generate a unified dilution ratio control signal through conflict judgment and strategy switching mechanism. In the current time window, according to the control state determined by the disturbance intensity and the control conflict, the corresponding control strategy combination and control execution cycle are selected. If it is in a low disturbance and low conflict state, the proportional integral derivative control strategy and the adaptive control strategy participate in adjustment together, the cycle is compressed to within 50% of the basic cycle, forming a high-frequency and high-precision control signal output path. Under this condition, based on the multi-dimensional parameter input of flow, conductivity, solid content rate, pressure and temperature, the two control strategies generate dilution ratio adjustment amounts respectively, and through conflict detection logic, the final dilution ratio control signal is synthesized for subsequent driving of the actuator operation.

[0072] The adaptive control algorithm in the present application can be realized by using existing mature control methods, such as model reference adaptive control (MRAC), gain adjustment type fuzzy adaptive control or self-learning control algorithm based on least mean square error estimation, or other control algorithms with dynamic parameter adjustment capability, which does not constitute a limitation to the protection scope. In one specific embodiment of the present application, the adaptive control strategy uses a model reference adaptive control method from the existing control engineering field, the core idea of which is: an ideal target response model is constructed, and the actual output behavior of the current system is continuously compared with the target model. If there is a response difference, the control parameters are automatically adjusted through error feedback, so as to dynamically optimize the adjustment behavior and ensure that the dilution ratio of fecal sewage liquid can stably approach the set solid content rate range.

[0073] The method first presets an expected system dynamic response model, for example, after a given injection of clean water, the slurry concentration should reach the set value within a certain time and be maintained within a certain fluctuation range. Subsequently, during the delivery of fecal sewage liquid, the system continuously observes the actual concentration change caused by the dilution ratio control result according to the real-time data collected by the sensor. The change trend is gradually compared with the target model output. When it is detected that the concentration change speed is slow, the overshoot amplitude is too large or the response delay is too large, the system will start the control parameter updating mechanism.

[0074] In this mechanism, the parameter weights used to generate control instructions are corrected in real time according to the size and trend of the error between the actual output and the target output. These weights can be understood as the "internal memory" of the controller summarizing the historical adjustment behavior to determine whether the water injection flow should be increased or decreased at the next moment, so as to make the control strategy approach the optimal one. This method has self-learning ability and can maintain high-precision mixing control performance under the interference of uncertain factors such as nonlinear disturbance, pump valve aging, and pipe resistance changes.

[0075] The adaptive control strategy in this embodiment does not rely on specific physical models or fixed control gains, but adjusts control decisions based on sensor feedback and errors. It has wide adaptability and robustness. The method has been widely used in industrial control, process regulation and complex fluid transportation, and belongs to the category of existing control algorithms known in the art and can be directly reused. The adaptive control strategy described above can be implemented through common industrial programming platforms, edge computing devices or programmable logic control devices, and the specific control parameters can be flexibly set according to the target solid content regulation accuracy requirements.

[0076] Based on the generated dilution ratio control signal, the control instructions for variable frequency pump frequency and electric valve opening are mapped. Taking the target dilution ratio of 1.8:1 as an example, the current manure liquid flow is 90 L / min, and about 50 L / min of clean water needs to be injected. The control logic converts this water injection requirement into the target frequency output of the variable frequency pump (for example, set to 62% of the maximum frequency), and matches the corresponding opening value (for example, 47.8%) according to the valve flow characteristic curve, to ensure that the target water injection flow is stably executed. This process takes into account the current system pressure difference, pump efficiency changes and pre-valve dynamic flow rate to achieve high-fidelity conversion of control signals to physical actions.

[0077] Further, in actual use, the deviation between the actual water injection behavior and the set target can be verified by collecting the execution feedback parameters of the frequency conversion pump and the electric valve in real time after the control instruction is output, and a closed-loop feedback path is constructed. If the feedback data in the current period shows that the clean water injection amount deviates from the expectation (for example, the set value is 50 L / min, and the measured value is only 42 L / min), the system will calculate the deviation value and feed it back to the control strategy in the next period, and dynamically correct the amplitude of the control signal. This error correction process is updated every 5 seconds, ensuring that the dilution effect can still be maintained within the set target solid content rate range under the influence of external disturbances or actuator response lags. After the clean water injection is completed, the fecal sludge and clean water are dynamically mixed in the main mixing pipeline, and a solid content rate sensor is arranged downstream of the mixing point for real-time monitoring. If the monitoring value enters the set target interval (for example, 5.8%-6.2%), the current control rhythm is maintained; if it deviates continuously, the control state reenters the identification process and updates the control strategy selection. The entire mixing control link is based on state-driven control output, actuator physical behavior feedback, and real-time sensing of solid content rate to build a closed-loop link, ensuring the accuracy, stability, and adaptive ability of the mixing process in a complex disturbance environment.

[0078] In an embodiment of the present application, considering that there are factors such as start-up delay, flow rate stabilization delay, and nonlinear response of the frequency conversion pump and the electric valve in the actual execution of the control instruction, the system presets the response curves of various actuators in the initialization stage to form a mapping model from the instruction to the actual flow rate. For example, the average response time of the frequency conversion pump from receiving the speed setting instruction to the actual flow rate stabilization output is one to two seconds, and the electric valve opening adjustment process has a lag of 300 to 500 milliseconds. The system estimates and compensates the time point of the expected control behavior according to such response characteristics, so that the dilution ratio signal output by the controller can take into account the trend of the actual water injection amount in advance, reducing the risk of closed-loop jitter.

[0079] When the pipeline is laid with sensors, the concentration sampling point is selected at a position downstream of the clean water injection inlet and far enough from the mixing stabilization section to avoid false concentration fluctuations caused by local disturbances affecting control judgment. At the same time, the system can be provided with a concentration trend judge, which does not immediately trigger the controller response when the concentration just changes, but waits for several time windows to confirm that the concentration has indeed changed due to the control instruction, so as to realize the identification and tolerance of the response delay in the controller, and improve the stability of the closed-loop judgment.

[0080] In order to avoid frequent command oscillation due to inaccurate delay identification, the system sets a closed-loop calibration period mechanism, and only triggers the dilution ratio recalculation when the concentration deviation exceeds the set range in a plurality of consecutive sampling periods. At the same time, if the current control state belongs to a high disturbance state, the system will reduce the control frequency and prolong the control period to prevent the control command update speed from causing closed-loop "rear-end" oscillation, thereby ensuring the stability of the closed loop.

[0081] The above algorithms or formulas are dimensionless values calculated, and the results of the latest real situation are obtained by collecting a large amount of data and software simulation. The preset parameters are set by the person skilled in the art according to the actual situation.

[0082] It should be understood that the size of the sequence number of the above processes in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0083] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. The person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0084] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-described device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0085] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for dynamically regulating the proportion of manure liquid based on sensory feedback, characterized in that, The method comprises the following steps: A plurality of sensors are arranged in the fecal liquid conveying pipeline to collect flow rate, conductivity, solid content, pressure and temperature data; A multi-dimensional parameter joint analysis method is used to calculate the load fluctuation entropy value in a set time window, which is used to represent the disturbance intensity and complexity of system operation; The proportional-integral-derivative control output and the adaptive control output are recorded synchronously, and the conflict frequency between the two types of control instructions is counted in the time window to generate a control conflict count value; According to the combination of the load fluctuation entropy value and the control conflict count value, the current control state is determined, and the control state is divided into four states: high disturbance and high conflict, high disturbance and low conflict, low disturbance and high conflict, and low disturbance and low conflict; According to the control state type, the priority and control execution frequency of the control strategy are dynamically adjusted. In the high disturbance and high conflict state, the single proportional-integral-derivative control is used and the frequency is reduced. In the low disturbance and low conflict state, the proportional-integral-derivative control and the adaptive control are allowed to run cooperatively and the frequency is increased. In the remaining states, the switching is performed according to the preset rules. A dilution ratio control signal is output to control the frequency conversion pump and the electric valve to inject clean water and mix the fecal liquid, thereby performing the mixing control within the target solid content range.

2. The method according to claim 1, wherein, When the sensors are used, a time-sharing synchronous acquisition mechanism is adopted to trigger the data reading of all sensors according to a unified time reference, to obtain a complete parameter group at the corresponding time point, and to pre-set a data validity determination rule to eliminate abnormal values and missing values generated in the acquisition, thereby forming a continuous parameter sequence that can be used for subsequent load fluctuation entropy value calculation.

3. The method according to claim 1, wherein, Before calculating the load fluctuation entropy value, the following processing steps are included: The collected flow rate, conductivity, solid content, pressure and temperature data are normalized to form a multi-dimensional parameter input vector. The multi-dimensional parameter input vector is segmented by a set time window to construct a multi-dimensional time sequence parameter matrix. The parameter matrix is filtered to remove noise and abnormal values to generate an effective data set for entropy value calculation.

4. The method according to claim 1, wherein, The calculation of the load fluctuation entropy value includes the following steps: A joint probability distribution model is established for the effective data set using the kernel density estimation method, and then the Shannon entropy value is calculated based on the joint probability model to obtain a unique load fluctuation entropy value. The refresh period and window update strategy of the entropy value output are set to update the load fluctuation entropy value in real time.

5. The method according to claim 4, wherein, The generation of the control conflict count value includes the following steps: The control instruction sequences output by the proportional-integral-derivative control strategy and the adaptive control strategy are recorded in the set time window. The control conflict determination standard is set. The directions and amplitudes of the two types of control instructions at the same time point are compared. When the adjustment directions are opposite and the amplitude difference exceeds the first conflict threshold, it is counted as a complete control conflict event. When the adjustment directions are consistent but the amplitude difference exceeds the second conflict threshold, the weighted conflict frequency is calculated according to the difference. All complete control conflict events and weighted conflict frequencies are accumulated in the time window to form the control conflict count value, which is dynamically updated with the time window sliding.

6. The method according to claim 5, wherein, The time window in the generation of the load fluctuation entropy value and the control conflict count value is consistent.

7. The method according to claim 6, wherein, The determination of the current control state includes the following steps: A disturbance threshold interval of the load fluctuation entropy value and a conflict threshold interval of the control conflict count value are set, and the entropy value and the count value are mapped to a disturbance level and a conflict level, respectively; A two-dimensional state mapping matrix is constructed, and the disturbance level and the conflict level are combined and correspond to four types of control states, wherein the disturbance level is divided into high disturbance and low disturbance, and the conflict level is divided into high conflict and low conflict; A state retention mechanism is used to make consistency judgment on the state results of continuous multiple time windows, and if the state remains unchanged for more than a set period, the state change is confirmed to be valid.

8. The method according to claim 7, wherein, When the determined control state does not last for a set confirmation period, a historical state record window containing the current time point is extracted, and the occurrence frequencies of four states, i.e., high disturbance high conflict, high disturbance low conflict, low disturbance high conflict and low disturbance low conflict, in the window are counted; The probability value of the current state to be confirmed in the historical state record window is calculated and compared with a preset probability threshold value; When the historical probability of the current state to be confirmed is greater than or equal to the preset probability threshold value, the current state to be confirmed is directly confirmed as the current control state, otherwise, the last confirmed state remains unchanged.

9. The method according to claim 8, wherein, The priority and control execution frequency of the control strategy are dynamically adjusted, including the following steps: A mapping relationship between the control state and the execution period is established, the high disturbance high conflict state is corresponded to the lowest control frequency interval, the low disturbance low conflict state is corresponded to the highest control frequency interval, and the specific control period values corresponding to each type of state are set; When the control state is high disturbance high conflict, the adaptive control strategy execution is suspended, and the proportional integral derivative control period is extended to more than twice the basic period; when the control state is low disturbance low conflict, the parallel calculation module of the proportional integral derivative control and the adaptive control is activated, and the control period is compressed to within 50% of the basic period, and the remaining states select the intermediate control period according to the mapping rule and switch according to the set strategy.

Images