Method for predicting health degree of injection pump

Abstract

The invention discloses a health degree prediction method for an injection pump. The method comprises the following steps: S1, recording operation historical data of the injection pump; s2, calculating the standard of the equipment in a normal operation state by using the operation historical data of the equipment; s3, respectively calculating an average value and a standard deviation of the equipment 5 minutes before the equipment is started, and adopting the average value and the standard deviation at corresponding moments during calculation and scoring; and S4, each index parameter is scored according to the weight, that is, four parameters are used, each parameter represents the health state of one subsystem, the highest score is 25, the lowest score is 0, and the sum of the four parameters is the equipment health degree. The invention belongs to the technical field of injection pump health prediction. The method has the technical effects that the real-time flow, the inlet and outlet pressure difference, the power and the pump shaft vibration amplitude value of the injection pump at the T + 1 moment are respectively predicted by using the health degree prediction method, the health degree of the injection pump at the T + 1 moment is calculated by the predicted value of each parameter, the calculation mode is referred, and the health degree evaluation mechanism is the health degree predicted value of the equipment at the T + 1 moment.

Description

Technical Field

[0001] The invention relates to the technical field of injection pump health prediction, and in particular to a method for predicting the health of an injection pump. Background Art

[0002] The injection pump is a machine that conveys fluids or pressurizes fluids. It is mainly used to convey water, oil, ore pulp, acid and alkali liquids, emulsions, suspensions, gas mixtures and liquid metals. It is a common conveying equipment in the mining, chemical and metallurgical industries. The teeth of the two gears of the gear pump are separated from each other, forming a low pressure, the liquid is sucked in, and sent to the other side along the shell wall. The two gears on the other side close together, forming a high pressure to discharge the liquid.

[0003] There are four main factors that affect the health of the injection pump, namely flow rate, inlet and outlet pressure difference, pump shaft vibration amplitude and real-time power.

[0004] It is not easy for existing injection pumps to predict the real-time flow rate, inlet and outlet pressure difference, power, and pump shaft vibration amplitude at time T+1. It is not possible to use the predicted values ​​of each parameter to calculate the health of the injection pump at time T+1, resulting in the inability to predict the health of the injection pump in real time. Summary of the invention

[0005] To this end, the present invention provides a method for predicting the health of an injection pump to solve the above-mentioned problems in the prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] According to a first aspect of the present invention, a method for predicting the health of an injection pump comprises the following steps:

[0008] S1: The injection pump is started for a period of time, and the historical data of the injection pump operation during this period is recorded;

[0009] S2: Calculate the mean and standard deviation of the injection pump historical data respectively;

[0010] S3: Each mean value and standard deviation is scored according to the weight, using 4 parameters. Each parameter represents a health status, and the sum of the 4 parameters is the health of the device.

[0011] Furthermore, the four parameters are flow rate, inlet and outlet pressure difference, pump shaft vibration amplitude and real-time power.

[0012] Furthermore, the calculation formula in S3 is:

[0013]

[0014] in It is the average value of various historical operation data.

[0015] Furthermore, the calculation formula in S3 is:

[0016]

[0017] Where σ is the standard deviation of each operating historical data.

[0018] Furthermore, the score calculation method has an average value of and a standard deviation of, so its score is:

[0019]

[0020] The next step of calculation is based on the specific number of p.

[0021] Furthermore, if p<=2, the score is:

[0022] S=m*(1-p / 2)

[0023] If p>2, the score is S=0;

[0024] S is the single parameter score, and m is the highest score of a single parameter.

[0025] Furthermore, the health prediction forms a cyclic unit. As the input operation history data continues to increase, the cyclic unit passes the previous state to the current input, and the current new input data is used to predict the health of the current round.

[0026] The present invention has the following advantages: a health prediction method is used to predict the real-time flow rate, inlet and outlet pressure difference, power, and pump shaft vibration amplitude of the injection pump at time T+1, and the health of the injection pump at time T+1 is calculated using the predicted value of each parameter (calculation method reference: health evaluation mechanism) as the predicted value of the health of the equipment at time T+1. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The process of the method for predicting the health of an injection pump provided by the present invention is as follows: Figure 1 .

[0028] Figure 2 The process of the method for predicting the health of an injection pump provided by the present invention is as follows: Figure 2 .

[0029] Figure 3 The process of the method for predicting the health of an injection pump provided by the present invention is as follows: Figure 3 .

[0030] Figure 4 The process of the method for predicting the health of an injection pump provided by the present invention is as follows: Figure 4 .

[0031] Figure 5 This is the inlet and outlet pressure fluctuation diagram provided by the present invention.

[0032] Figure 6 This is a fluctuation diagram of power consumption of the injection pump provided by the present invention after starting for 30 minutes.

[0033] Figure 7 The power consumption fluctuation diagram after the temperature provided by the present invention.

[0034] Figure 8 This is the basic structure diagram of the RNN provided by the present invention.

[0035] Fig. 9 The present invention provides three RNN diagrams with continuous loop structures.

[0036] Fig.10 LSTM diagrams of three continuous loop structures provided by the present invention. DETAILED DESCRIPTION

[0037] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Example 1

[0039] like Figures 1 to 10 As shown, a method for predicting the health of an injection pump in an embodiment of the first aspect of the present invention comprises the following steps:

[0040] S1: The injection pump is started for a period of time, and the historical data of the injection pump operation during this period is recorded;

[0041] S2: Calculate the mean and standard deviation of the injection pump historical data respectively;

[0042] S3: Each mean value and standard deviation is scored according to the weight, using 4 parameters, each parameter represents a health status, and the sum of the 4 parameters is the health of the device;

[0043] The highest score is 25 points and the lowest is 0 points. The sum of the four parameters is the health of the device.

[0044] In the above embodiment, it should be noted that the injection pump is started for a period of time generally of 5 minutes, and the health evaluation of the injection pump uses parameter indicators:

[0045] (1) Flow rate: The function of the injection pump is to inject liquid CO2 into the ground. The amount of CO2 injected represents its effect and also reflects its working efficiency.

[0046] (2) Inlet and outlet pressure difference: When CO2 passes through the injection pump, the pressure determines the effect of CO2 injection and also reflects the operating status of the pump;

[0047] (3) Pump shaft vibration amplitude: The vibration amplitude of the pump's main motor represents the condition of shaft wear;

[0048] (4) Real-time power: The power consumed by the injection pump during operation, i.e., current multiplied by voltage.

[0049] Example 2

[0050] like Figures 1 to 4 As shown, a method for predicting the health of an injection pump includes all the contents of Example 1. In addition, the four parameters are flow rate, inlet and outlet pressure difference, pump shaft vibration amplitude and real-time power.

[0051] Example 3

[0052] like Figures 1 to 4 As shown, a method for predicting the health of an injection pump includes all the contents of Example 2. In addition, the calculation formula in S3 is:

[0053]

[0054] in It is the average value of various historical operation data.

[0055] Example 4

[0056] like Figures 1 to 4 As shown, a method for predicting the health of an injection pump includes all the contents of Example 3. In addition, the calculation formula in S3 is:

[0057]

[0058] Where σ is the standard deviation of each operating historical data.

[0059] After the RNN is expanded, it is connected by multiple identical units. However, the actual structure of the RNN is a self-circulating structure as shown on the left side of the figure above. That is, as the input data continues to increase, the self-circulating structure passes the previous state to the current input, and together with it, it is used as the new input data for the current round of training and learning until the input or training ends. The final output is the final prediction result.

[0060] LSTM is a special type of RNN. The difference between the two is that a common RNN has only one state in a single loop structure. However, a single loop structure (also called a cell) of LSTM has four states. Compared with RNN, the LSTM loop structure maintains a persistent cell state and continuously passes it on to decide which information to forget or continue to pass on.

[0061] Example 5

[0062] like Figures 1 to 4 As shown, a method for predicting the health of an injection pump includes all the contents of Example 4. In addition, the scoring calculation method has an average value of and a standard deviation of, and its score is:

[0063]

[0064] The next step of calculation is based on the specific number of p;

[0065] If p<=2, the score is:

[0066] S=m*(1-p / 2)

[0067] If p>2, the score is S=0;

[0068] S is the single parameter score, and m is the highest score of a single parameter.

[0069] In the above embodiment, it should be noted that a persistent unit state is maintained between the cyclic units and continuously passed on to determine which information to forget or continue to pass on;

[0070] like Figure 5 As shown, each line represents the pressure change curve after the equipment is started once. The horizontal axis is the time axis after the start-up, that is, the 1st, 2nd, 3rd... minute, and the vertical axis is the corresponding pressure. As shown in the figure above, the average value and standard deviation are calculated for the first 5 minutes of the equipment startup. The average value and standard deviation at the corresponding time are used when calculating the score because of its special state at the time of startup; from 6 minutes to the end, an average value and standard deviation are calculated because it has been in a stable and normal operating state.

[0071] Example 6

[0072] like Figures 1 to 4 As shown, a method for predicting the health of an injection pump includes all the contents of Example 5. In addition, the health prediction forms a cycle unit. As the input operation history data continues to increase, the cycle unit transfers the previous state to the current input, and the current new input data is used to predict the health of the current round;

[0073] In the above embodiment, it should be noted that the number of times a single recurrent unit needs to self-update is determined by the dimension of the input data and the number of cycles, rather than by the connection of multiple single recurrent units;

[0074] The technical effects achieved by the above embodiments are as follows: the health prediction cycle unit is composed of an input gate, a forget gate, an output gate and a unit state;

[0075] Input gate: determines how much of the network's input data needs to be saved in the unit state at the current moment.

[0076] Forget gate: determines how much of the unit state at the previous moment needs to be retained to the current moment.

[0077] Output gate: controls how much of the current cell state needs to be output to the current output value.

[0078] like Figure 1 As shown in the figure, the output h_t-1 at the previous moment and the current data input x_t are used to obtain f_t through the forget gate:

[0079] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0080] like Figure 2 As shown in the figure, the output h_t-1 at the previous moment and the current data input x_t are used to obtain i_t through the input gate, and the temporary state C~t at the current moment is obtained through the unit state:

[0081]

[0082] like Figure 3 As shown, the process of obtaining the current cell state C_t from the previous cell state C_t-1, the forget gate output f_t, the input gate output i_t and the cell state output C~t:

[0083] C t =f t *C t-1 +i t *C t ;

[0084] like Figure 4 As shown, the output h_t-1 at the previous moment and the current data input x_t, the process of obtaining o_t through the output gate, and the final output is obtained by combining the current cell state C_t and o_t

[0085] h_t process: h t =o t*tanh(C t );

[0086] A layer of LSTM is composed of a single loop structure. The dimension of the input data and the number of loops determine how many times a single loop structure needs to update itself, rather than multiple single loop structures connected together. That is, the total number of parameters of the current layer of LSTM only needs to be calculated for one loop unit, rather than the total number of multiple consecutive units.

[0087] The following is a set of diagrams to explain the basic composition and implementation principle of LSTM cells in detail. LSTM cells consist of input gate, forget gate, output gate and cell state.

Claims

1. A method for predicting the health of an injection pump, characterized in that: The steps include: S1: The injection pump is started for a period of time, and the historical data of the injection pump operation during this period is recorded; S2: Calculate the mean and standard deviation of the injection pump historical data respectively; S3: Each mean value and standard deviation is scored according to the weight, using 4 parameters. Each parameter represents a health status, and the sum of the 4 parameters is the health of the device.

2. The method for predicting the health of an injection pump according to claim 1, characterized in that: The four parameters are flow rate, inlet and outlet pressure difference, pump shaft vibration amplitude and real-time power.

3. The method for predicting the health of an injection pump according to claim 1, characterized in that: The calculation formula in S3 is: in It is the average value of various historical operation data.

4. A method for predicting the health of an injection pump according to claim 3, characterized in that: The calculation formula in S3 is: Where σ is the standard deviation of each operating historical data.

5. The method for predicting the health of an injection pump according to claim 4, characterized in that: The scoring calculation method is the mean value and the standard deviation, so its score is: The next step of calculation is based on the specific number of p.

6. A method for predicting the health of an injection pump according to claim 5, characterized in that: If p<=2, the score is: S=m*(1-p / 2) If p>2, the score is S=0; S is the single parameter score, and m is the highest score of a single parameter.

7. The method for predicting the health of an injection pump according to claim 1, characterized in that: The health prediction forms a cyclic unit. As the input operation history data continues to increase, the cyclic unit passes the previous state to the current input.

8. The method for predicting the health of an injection pump according to claim 7, characterized in that: The new input data is used to predict the health of the current round.

Images