Plant fault diagnosis system

~9~985 PLANT FAUI.T DIAGNOSIS SYSTE:M BACKGROUND OF T~E INVE:NTION 1.

Filed of the Invention This invention relates to a plant fault diagnosis system capable of diagnosing fault or failures and applied to a large scale plant such as chemical plants.

2.

Description of the Related Art Hitherto, a diagnosis system for diagnosing faults and failures generated at plants has a function capable of detecting a fault generated in a process of the plant and provided with a predetermined threshold upon the process value, whereby a fact that the process value deviates from the threshold is notified as an alarm to operators.

However, with the above-described type of method, when a fault is generated in a process and the disturbance of the process is propagated, the number of alarms increase. Therefore, it becomes very difficult for the operator to acknowledge the type of the fault generated in the process and the cause thereof. To this end, as means to overcome the above-described type of problem, a method is known in which the number of alarms are reduced by displaying the logical product of a plurality of these alarms which are considered to represent the same information as the representative '" ~z~ -2~ value of the alarms. Another known method is such that the way to display the critical alarm is distinguished from general alarms.

However, according to the above-described technology, even if any fault is detected in a process, only detected is the disturbance generated in the process. Therefore, the cause for the faults cannot be identified. A method can be considered to identify the cause of a fault by a pattern matching way such that the alarm items expected to be generated in accordance with the causes for faults are previously picked up and the logical product thereof is used. However, with this method, all of the alarms need to be issued in the same form to identify the cause. That is, a problem arises that the cause cannot be detected if this type of alarm cannot be detected in a case where the pattern matching method is applied to the alarm of a poor probability to be detected such as an alarm representing the disturbance degree of the process value is too small and a certain time delay is expected to occur from generation of the fault to detection of it.

A plurality of instruments such as pressure gauges and differential pressure gauges provided for chemical plants or the like are critical instruments for acknowledging the state of the operating condition of the plants. Although these instruments need to always work accurately, they can generate erroneous indications due ~29~S --3-to an indication level hold and a deviation after they have been used for a long time period. Hitherto, since means to automatically detect such fault of the instruments have not been available, it needs to depend upon the detecting means of human power of the operators.

That is, the operator performs a fault detection by way of detecting the erroneous indication of the instruments after they have checked the indicated levels by the instruments from the fault state through the plant, or after they have made a comparison of the indicated ~evels to those of the field instruments.

However, a problem arises that a quick and proper fault detection cannot be performed if such detection depends upon the operator's human powerO As a result of this, it is feared that, for example, although the pressure of a plant is at the abnormal level, the erroneous indication shown on the instrument causes the operator to determine that the pressure is in the normal range. It leads to a fact that an early detection of the faults in the plant cannot be per~ormed. On the contrary, iX i3 as weIl ~e~r~ that although the pressure is in the normal range, the erroneous indication shown on the instrument cau5es the determination o the abnormal pressure. As a result of this, there is the possibllity of performing erroneous operation.

As a means for diagnosing fault in the plant.

there are off-line fault diagnosing systems in which ~z~7sa8~;

fault diagnoslng is performed by input-tiny process data of the plant in off-line-input, and on-line fault diagnosing system in which the data is on-line-input.

Process data to be input to the types of diagnosing systems are, without exception, measured by the corresponding instruments. In an off-line system, the result of field reading the indication is manually input, while in an online system, data is input through signal lines. Howéver, for example, in chemical plants, the terminal end of the detecting portion of the instrument may be corroded due to the characteristics of the subject to be treated, or clogging can accrue. In the instrurnents having movable portions, the movable portions can be snagged or the like, causing normal working to be prevented.

As described above, in large sized plants such as chemical plants, if the fault diagnosing is perEormed by using the process value of the damaged instruments, fault generation can be detected without any fault being generated, causing an alarm or the like to be erroneously issued. Such erroneous indications cau~e operators to become needlessly confused, and is troublesome when operating plants.

At the time of performing plant instrumentation, the upper limit and the lower limit of the normal operation range with respect to each of the processes are provided and a device for issuing an alarm ~9~85 when the process value deviates from the normal operation limi~s is installed for the purpose o~ preventing breakdowns due to faults at the time of instrumentation.

When a fault is generated in the instruments of a chemical plant, the local disturbance successively propagates to other portions widely. As a result of this, only one failure can affect many process values. The instrument needs to identify the cause of the generation of the fault so as to remove it, but it is also important to maintain each of the disturbed process values within a safety range. The reason for this is that entering the disturbed process value due to the original failure into the unfavorable range can cause another failure. If a multiplicity of process value disturbances are simultaneously generated, it needs to select the disturbance of the process value capable of causing a critical failure from a large number of alarms so as to preferentially act to overcome the critical failure.

Therefore, in the instrumentation of chemical plants, systems were disclosed with which an alarm is issued promptly for notifying the generation of a Eault by performing a faulk diagnosis with computers. ~n these alarm systems, a plurality of alarms are issued for each process to which the disturbances are propagated and, in order to select the process value capable of causing a critical failure from each of the process ~L~9~15 --6-values which have generated disturbances, a classification is employed such as that, for example, the color of the alarm displaying lamp is classified in accordance with the degree of importance factor.

However, the alarm systems of the type as described above confront the following problems.

That is, such systems notify a fact that an abnormal state was generated in the instrumentation of the chemical plant after this state has occurred, but these systems cannot issue alarms previously by detecting a ~act that the process value is now deviating from a normal range. Therefore, it arises a problem that an action to overcome the abnormality in a chemical plant is caused to be too much delayed.

On the other hand, a method can be employed in which the upper and lower limits of the process value are provided in a further safety range so that a fault is detected in an early stage. However, there is the probability of issuing an erroneous information depending upon a determination of fault although the state is normal. I Furthermore, the alarms are selected only on the basis that the importance factor of the fault of the chemical plant generated rom deviation from a normal range of the process value, but selection of issue of the alarms in accordance with a floating time (marginal time) taken to reach the generation of the fault is not 8~ --7-considered. As a result of this, although there is insufficient floating time to generation of a fault, and although the importance factor of such fault is critical, no alarm is issued, causing a fault.

Furthermore, a conventional plant fault diagnosing system is known in which the physical quantity of each of the portion in a plant is always measured with sensors for the purpose of issuing an alarm when the physical quantity or its change rate is excessively different from the value at the normal operation sate of the plant.

However, in these type of plant fault diagnosing systems confront the following problems.

.. Since the allowable range for the physical quantity is arranged such that the physical quantity can be present applicable to all of the operation conditions of the plant, the allowable range needs to be large. In such case, a certain long time is ta~en from generation of a fault in the plant to issue of an alarm, and it becomes a cause for the successive generat.ion of faults in each of the portions in the plant. As a result of this, an excessive long time period is taken to recover the fault. .

When an alarm is set for the change rate, such setting can cause alarm issue depend.ing upon a determination of generation of the fault although the state cannot actually be determined that the fault \ occurred. That is, an alarm can be issued relatively earlier from the start of the physical quantity change due to generation of the fault, however, the alarm is obliged to be issued even if the measured value contains a disturbance such as noise or when the physical quantity changes slightly.

Furthermore, since the number of points at which the physical quantity is measured is large in the plant instrumentation, it takes too much labor for operators to completely and manually determine the allowable range of the fluctuation with respect to the value at the time of normal operation.

SUMM~RY OF T~ INVENTION .. . A fir.st object of the present... invention is to provide a plant fault diagnosing system capable of notifying operators the types and causes of the generated fault when it occurs.

In order to achieve the above-described ob;ect, a preerred embodiment comprising: a knowledge memory for storing information for detecting faults or the like for storlng the corresponding relationship between a hypothesis that a Eailure or a Eault sate is generated and symptoms caused to appear when the hypothesis is occured in such a manner that the corresponding relationship is classified depending upon the possibility of appearance of the symptom corresponding to each hypothesis or intensity of the detection of the same and 9 symptoms which cannot be appeared: a diagnosing portion to which data representing the symptoms which are now being appeared in the plant is input and capable of determining whether each hypothesis is occured or not by collating the data with the contents written in the knowledge memory; and notifying means for notifying the diagnosis result from this diagnosing portion, whereby the type and the cause of khe fault can be displayed to operators so that even if the appearance possibility of the symptoms are different, a proper determination can be performed in accordance with the classification of the symptoms.

A second object of the present invention is to provide a .plant ~ault diagnosing system capable of informing generation of an instrument failure and preventing an erroneous information due to the failure for the purpose of protecting operators from needless confusion.

In order to achieve the second object, the preferred embodiment comprises: an instrument failure diagnosing portion for diagnosing a failure of each of instrument, the instrument fallure being diagnosed on the process data in the plant; a general ault diagnosing portion for diagnosing abnormalities other than the failures of the instrument, the general fault being diagnosed on the result in the instrument failure diagnosing portion and the process data; and a notifying - 1 0 portion for notifying the diagnosis result of the general fault diagnosing portion and the diagnosis result of the instrument failure diagnosing portion, wherein the general fault diagnosing portion displays the failure when the failure is detected by the instrument failure diagnosing portion, and as well performing abnormal diagnosis by removing the process values of the instrument in which the failure is detected from the process data, whereby the failure is notified when the failure of the instrument is detected, and abnormal diagnosis is performed by removing the process value of the subject instrument from the process data so that needless confusion of the operators can be avoided by preventing the erroneous information due to the failure of the instrument.

A third object of the present invention is to provide a plant fault diagnosing system capable of issuing an alarm before occurrence of an abnormality by detecting a fact that the process value in an instrumentation of a chemical plant is now deviàting from a normal range, and as well capable of issuing alarms successively in accordance with a fault having poor floating time or in accordance with the urgency factor degree considering floating time and importance factor.

In order to achieve the third object, a preferred embodiment comprising: a memory for storing the importance factor when each alarm depending upon alarm - 1 1 factors included in information for detecting faults or the like and also for vutputting the importance factor to the diagnosing portion is juxtaposed to the diagnosing portion, wherein the diagnosing portion including means for estimating the time taken to generation of each abnormal state depending upon the physical quantity of the data, means for estimating the time required to avoid an abnormal state, means for calculating an estimated floating time required for avoidance operation in such a manner that an estimated avoidance required time is subtracted from an estimated abnormality generation time obtained by the means, and means for calculating urgency factor using the estimated floating time and the importance factor, whereby the notifying means is constituted in such a manner that it is capable of ranking and displaying the abnormalities in the sequential ordsr upon the intensity oP the urgency factor, and as well ranking and displaying the same in the sequential order upon the floating time. As a result of this, when any abnormality is generated in the plant, the faults having high urgency factor and having short floating time can be taken a measure required in each portion of the plant.

A fourth object of the present invention is to provide a plant fault diagnosing system capable of early detecting a fault in accordance with the state where the plant is operated, and from which a reliable alarm -12without limited erroneous information can be obtained.

In order to achieve the four-th object, a preferred embodiment is characterized in that the changeable reference value comprises an average value obtained by using an average calculator which is juxtaposed to the diagnosing portion, this average value being obtained at a predetermined time interval of the physical quantity at specific points in the plant which is the data, wherein the notifying means issues an alarm when the difference between the average value and the physical quantity deviates from a predetermined allowable range, and the average value is arranged not to be renewed during a period where the alarm is being issued.

As a result of this, even if the value of the physical quantity is slightly changed due to the conditions such as outside temperature or the like, a stable value corresponding to the process change through a long time period can be determined by re-calculating the average value at a proper time period intervals.

BRI~F DFSCRIPTION OF TH~ DRAWINGS Fig. 1 is a block diagram illustrating a ~irst embodiment o~ the present lnvention;

Fig. 2 i9 a block diagram illustrating a modi~ied example of the former;

Fig. 3 is a view illustrating a specific applied example of the above-described embodiment;

Fig. 4 is a view illustrating a classi~ication ~Zg7g~ criterion of a knowledge memory;

Fig. S is a view illustrating the relationship between hypothesis stored in the knowledge memory and the state of the appeared symptom;

Fig. 6 to 14 illustrate knowledge examples for each hypothesis;

Fig. 15 illustrates a process flow of the diagnosing portion of the first embodiment;

Figs~ 16 illustrates a process flow of the diagnosing portion of the above-described modified example;

Fig. 17 illustrates a block diagram according to a second embodiment;

Fig. 18 illustrates a displayed example of the failure points and the diagnosis result according to the second embodiment;

Figs. 19 to 22 illustrate the basic system structure according to a third embodiment;

Figs. 23 to 24 each illustrate different examples in which the above-described system is applied to the instrumentation of a chemical plant;

Fig. 25 illustrates a classification criterion of importance ~actor memory;

Fig. 26 illustrates the degree of urgency factor;

Fig. 27 illustrates the structure according to a fourth embodiment; and ~ 7~ --1 ~ Figs. 28 to 31 each illustrate modiEied examples of the fourth embodiment.

DETAILED DESCRIPTION OF TIIE PREFERRED EMBODIMENT Referring to Fig. 1 showing a first embodiment of the present invention, reference numeral 1 represents a knowledge memory which stores in~ormation for detecting such as faults for storing the corresponding relationship between a hypothesis that a failure or a fault state is generated in a plant and a symptom caused to appear when the hypothesis is generated in such a manner that the corresponding relationship is classified to the depending upon the possibility of appearance of the symptom corresponding to each of the hypothesis or intensity of the detection of the same and the symptoms which can not be appeared. Reference numeral 2 represents a diagnosing portion for performing a diagnosis whether each hypothesis actually is generated or not by way of inputting diagnosing patterns corresponding to the nowappearing fault state, these diagnosing patterns being able to be obtained in accordance with the process data supplied from each of the instruments in the plant. The diagnosing portion performs the diagnosis by way of further performing collation of the thus-input symptom appeared state, information on the non-appeared symptom, the relationship between the hypothesis stored in the knowledge memory 1 and the appeared symptom state, and the non-appeared symptom. Reference numeral 3 represents -15an ou~put display portion to serve as means ~or noti~ying the diagnosis result o~ the diagnosing portion 2.

AS the diagnosis logic in the diagnosing portion 2, a variety of information industrial methods such as that depending upon the certainty factor can be employed. The description "appearance and detection possibility" in the classification criterion in the knowledge memory 1 is an index representing a fact that the magnitude of the disturbance of the process value during generation of the fault is large enough to be detected, or the. symptom is suitable to be used for diagnosis slnce any.long time delay or the like cannot.be generated from generation of the fault to the appearance o~ the symptom. For examyle, the symptom classified to have a high appearance or detection possibility is connected, by way of being detected, to the determination in the diagnosing portion 2 that "a corresponding hypothesis is generated", while the same is connected to the determination in the diagnosing portion 2 that "corresponding hypothesis is not generated".

The symptom classi~ied to have a low appearance or detection possibility is connected to, by way of being detected, the determination in the diagnosing portion 2 that "corresponding hypothesis is generated". However, even i~ the same is not detected, it cannot be connected to the determination in the dlagnosing portion 2 that "corresponding hypothesis is not generated".

~9~g~ -16 In the classification criterion in the knowledge memory 1, the description "non-appeared symptom" represents a fact that appearance of such symptom due to the effect of the hypothesis cannot be considered from the physical properties of the plantO The symptom classified to be a non-appeared symptom state is connected to, by way of being detected, the determination the diagnosing portion 2 that "corresponding hypothesis is not generated". ..

As described above, the relationship between a hypothesis that a ~ailure or a fault is generated in the plant and the symptom appearing at the time of generation ..

of the hypothesis is classified in accordance with the magnitude of appearance or the detection possibility with respect to each hypothesis and the non~appeared symptoms so as to have it stored in the knowledge memory 1. On the other hand, the symptom which is now appearing in the plant is input to the diagnosing portion 2 50 as to perform diagnosis by collation with the contents written in the knowledge memory 1 and perform determination whether each hypothesls is generated or not. As a result of this, the type and the cause of the generation of the fault can be informed the operator so that the operator can properly determine in accordance with the classification of the symptom even if the appearance possibility is different.

Next, an modified example of the above 85i -17described first embodiment will now be described with reference to Fig~ 2. Referring to this figure, reference numeral 4 represents an knowledge memory, and reference numeral 5 represents a diagnosing portion. The portions which are the same as those shown in Fig. 1 are given the same reference numerals and the description for the same is omitted. In order to distinguish the knowledge memory 1 and the diagnosing portion 2 in the above-described embodiment from the knowledge memory 4 and the diagnosing portion S in this modified example, the knowledge memory 1 and the diagnosing portion 2 in the above-described embodiment are arranged to be a first knowledge memory and a first diagnosing portion, respectively, while the knowledge memory 4 and the diagnosing portion 5 in this modified example are arranged to be a second knowledge memory and a second diagnosing portion, respectively.

The second knowledge memory 4 according to this modified example stores the xelationship between the hypothesis that a failure or fault is generated in the plant and the symptom appeared at the time of generation of the hypothesis, the relationship being classified in accordance with the plant conditions. The diagnosing po~tion 5 collates the written contents in the second knowledge memory 4, selects the most suitable relationship from the written contents, and makes a comparison with the symptom which is now appeared in the plant, whereby the fault diagnosis is performed.

~7~8~ -18 There~ore, with this modified example, the cause of the fault can be properly displayed even if the symptom pattern at the time of ~ault generation is changed in accordance with the conditionsO Fig. 3 illustrates a liquid level control system for a cham~er used in chemical plants. Next, a case where the first embodiment is applied to the plant shown in this figure will be described. Referring to Fig~ 3, reference numeral 6 represents an inflow pipe, reference numeral 7 represents an inflow indicator for measuring the flow rate of the fluid passing through the inflow pipe.6. Reference numeral 8 represents an inflow control valve for controlling the flow rate of the f].uid passin.g. the inflow pipe 6, and reference numeral 9 represents a chamber for accumulating the fluid passing through the inflow pipe 6. Reference numeral 10 represents a level controller for measuring the fluid level in the chamher 9 and outputting a control signal to the inflow control valve 8 so as to secure the fluid level at the predetermined level. Reference numeral 11 represents a predetermined value set ~or the level controller 10. Reference numeral 12 represents an discharge pipe for taking out the fluid in the chamber 9.

Reference numeral 13 represents a control valve for controlling the flow rate o~ the fluid passing through the discharge pipe 12. Reference numeral 14 represents a discharge indicator for measuring the flow rate in the -1 9discharge pipe 12 and outputting a control signal to the control valve 13 so as to secure the flow rate at a predetermined value, and reference numeral 15 represents a set value for the discharge indicator 14.

In the thus-constituted plant, the following symptoms can be detected by using various conventional instrumentation systems and computers:

I.The level indicated for the inflow indicator 7 is high.

II.The level indicated for the inflow indicator 7 is low.

III.The level indicated for the level controller 10 is high.

IV.The level indicated for the level controller 10 is low.

V.The control signal for the level controller 10 is high (in this casel the inflow control valve 8 i9 operated to the open side). '.

VI. The control signal or the level controller 10 is low ~in this case, the inflow control valve 8 is operated to the close side).

VII.The value subtracting the indicated level for ~L2~7~8S -20the level controller 10 from the level estimated value is high~ wherein, the "level estimated value" is the level estimated in the chamber 9 by time-integrating the difference between the indicated level for the inflow indicator 7 and the indicated level for the discharge indicator 14 so as to estimate the retained quantity of the fluid in the chamber 9 so khat the level in the chamber 9 corresponding to the retained fluid quantity depending upon the geometric shape of the chamber.9.

This "level estimated value" can be expressed by the following equation:

Lg(t) = LM (~ (FI(t) - FC(t~dt ~ M ) .... (1) T-l T-l wherein, Lg(t) : level estimated value t : time FI(t) : indicated level for inflow indicator 7 FC(t) : indicated level for disaharge indicator 14 MT_I : level of the retained flu:Ld quantity in the chamber 9 at the time T ~ : integrating time T : present time LM : function (monotone increasing function) for obkaining the level in the chamber 9 from the retained fluid quantity in the chamber 9 and its -21geometric shape.

VIII. Indication of level for the level controller 10 subtracting from the level estimated value is low.

IX. Indication of level for the inflow indicator 7 subtracting from the inflow estimated value is high, wherein the "inflow estimated value" is a value obtained by estimation from the physical relationship between the open degree of the inflow control valve 8 and the flow rate oE the fluid passing through the inflow pipe 6, and the open degree of the inflow control valve 8 is replaced by a manipulated signal for the level controller 10. The inflow estimated value can be expressed by the following equation:

FIg = Cvi(Mil ~ ... (2) wherein FIg : inflow estimated value Mi : manipulated signal for operatlng the inflow control valve 8 Cvi(Mi): flow rate coe~ficient of the inflow control valve 8 ~the flow rate coefficient becomes the function of the operating signal, and is a monotone increasing function) dPi : the pressure difference of the flow inflow control valve 8 (it is assumed to be constant, here).

X. Indication of level for the inflow indicator 7 subtracting from the inflow astimated value is low.

The above-described symptoms I to X are shown as an example, and the items are not limited to one meaning. A variety of symptoms can be selected as needed in accordance with the method of writing knowledge to be described hereinafter.

In this embodiment, the hypothesis of the cause of the fault subjected to diagnosis is assumed. as follows:

~. high erroneous value indicated for the inflow indicator 7 ( higher value is indicated than the actual quantity passing through the inflow pipe 6) ~ low erroneous value indicated for the inflow indicator 7 (lower value is indicated than the actual quantity passing through the inflow pipe 6) inflow pipe 6 is clogged ~ high erroneous value indicated for the level controller 10 (hlgher value is indicated than the actual fluld level) ~ low erroneous value indicated for the level controller 10 (lower value ls indicated than the actual fluid level) ~ leakage from the chamber 9.

~gl7~ -23 The classification of the symptoms in accordance with -the appeared state is determined as shown in Fig. 4, and knowledge corresponding to the appeared symptoms due to generation of faults and its grounds in the first knowledge memory 1 about the hypothesis will now be described. In the following description, each hypothesis will be successively described by using the classification symbols and the symptom numbers shown in Fig. 4. However, the non-appeared symptom classification D includes the symptoms other than the classification symbols A, B, and C. It is, of course employed, that other symptoms can be classified in D by the limited numbers in order to distinguish other hypothesis.

Knowledge obtained as a result is shown in Fig. 5, and the knowledge examples in the hypothesis ~ to ~ are respectively shown in Figs. 6 to 10.

Knowledge corresponding relationship to symptom appearance due to generation of faults in the abovedescribed knowledge memory 4 and its ground will now be described. In the second knowledge memory ~, the written contents in the first knowledge memory 1 are stored after they have been classified in accordance with the various conditions. In the subject plant shown in Fig. 3, as the condition, two conditions are assumed, the two conditions being such that whether control o~ the level controller 10 is per~ormed or its control is released.

The ]cnowledge in a case where the control of the level ~2~79~ -24controller 10 is performed is shown in Figs. 6 to 10.

Therefore, the knowledge in a case where the control of the level controller 10 will now be described. The knowledge when the control of the level controller 10 obtained as a result is shown in Fig. 11, and examples of knowledge for the hypothesis ~ ~ ~ are shown in Figs.

12 to 14, respectively.

Figs. 15 shows a treatment flow when the treatment in the first diagnosing portion 2 is performed with a computer. Referring to this figure, reference numeral 16 represent a certainty factor determining logic for each hypothesis. In the first embodiment, the first diagnosing portion 2 performs fault diagnosis by calculating the certainty ~actor that a hypothesis is generated. In a case where the certainty factor is 100 or more is arranged to be the case where hypothesis generation is determined, while the same is arranged to be a candidate when it is less than this value. The more the certainty factor is large, the more the probability of generation of the hypothesis increases. When the certainty becomes zero, the hypothesis is considered to be completely abolished. In Fig. 15, the first diagnosing portion 2 performs the following treatment upon the all of the hypothesis stored in the knowledge memory 1:

(a) If one or more symptoms defined as the classification symbol D is appeared, the certainty factor 98~i -~5of the subject hypothesis is ~ero, ~b) if any symptom de:Eined as the classification symbol A is not appeared, the certainty factor of the hypothesis is zero, and (c) if a hypothesis in which the certainty factor is not zero after through the treatments (a) and (b) but all of the symptoms defined as the classification symbol B are appeared, its certainty is 100. In a case where there is non-appeared symptoms, the certainty factor is obtained by subtracting the number of the non-appeared symptoms from 100, and the calculation result becomes less than zero, its certainty factor is zero, (d) in the hypothesis in which the certainty factor of the hypothesis is 100 after through the abovedescribed treatments (a) to (c), if there is a symptom defined as the classification symbol C, its certainty factor is increased by 10.

Ater the treatment (a) has been performed, the hypothesis in which "irrational appearance of symptom" classified in classification D is appeared is abolished.

The hypothesis in which "symptom necessarily to be appeared" classified in classification ~ is not generated ater through the treatment (b) is abolished. When all of the symptoms considered to be appeared classified in classification B in the treatment (c) are generated, the hypothesis is determined to be correct. As the number of the appeared symptoms is reduced by one, its certainty -26factor is decreased. When the hypothesis which has been resulted to be generation by the action of the treatment (c) confronts the appearance of the "symptom WhiC]l are not necessarily detecte~, but its affection can be considered due to the structure of th~ process", generation of the hypothesis can be resulted with an intense certainty factor by the treatment (d) by using the certainty factor.

Fig. 16 illustrates the flow when the operation of the second diagnosing portion 5 is treated by a computer. In the second diagnosing portion 5, the relationship between the hypothesis stored in the second knowledge memory 4 and the symptoms which correspond to the condition is selected, and the certainty factor of each of the hypothesis is calculated in a manner similar to the first diagnosing portion 2.

Although in the above-described description, the first and second knowledge memories and the first and second diagnosing portions are distinguished at use, only one knowledge memory and one diagnosing portion can display the similar result at the output display portion depending upon the aapacity thereof.

The method of determining the certainty factor is not limited to the above-described embodiments, and other methods can be employed. Specifically, a method can be employed that is so arranged that a hypothesis is considered to be materialized if ten or more symptoms defined as the classification symbols A and B are appeared. Another method can be also employed that is so arranged that a hypothesis is considered to be materialized if the number of the non-appeared symptoms is less than 2 and simultaneously the symptoms defined as the classification symbols A and B exceeds 10.

Although, the description is made upon the case where the four classification symbols A to D are used, the number of the same may be less or more than four.

According to the first embodiment, the type and the cause of the generation of a fault can be displayed to an operator so that the operator can properly perform a judge in accordance with the classification of the symptom even if the symptom appearance possibility is d i f f erent.

Fig. 17 is a block diagram illustrating a second embodiment of the present invention. Referring to this figure, reference numeral 41 represents a data collection portion, reference numeral 42 represents instrument failure diagnose portion consistirlg a diagnosing portion. Reference numeral 43 represents a general fault diagnose portion as well consisting the diagnosing portion, and reference numeral 4~ represents a diagnosis result output display portion serving as an notifying means.

The above descri.bed data collection poxtion 41 collects process data from each of the instruments in the -28plant. The process data collected by this data collection portion 41 is input to the instrument failure diagnose portion 42 and the general fault diagnose portion 43. The instrument failure diagnose portion 42 diagnoses a fact whether there is a ~ailure in the instrument or the like by using the process data collected by the data collection portion 41, and it outputs its diagnosis result upon the failure point to the general fault diagnose portion 43 and the diagnos.is result output display portion 44~ As a method of diagnosing the instrument failure, a method can be ex.emplified by a method in which the .process data estimated from the material balance of the plant or heat balance is made comparison to the measured.result by the instrument, or another method can be exemplified that such failure is diagnosed when the slight change of the measured value becomes zero through monitoring the measured values.

The general fault diagnose portion 43 diagnoses a fault other than the instrument failure by using the process data input from the data collection portion 41..

It performs diagnosis by removing the process data for the instrument detected as failure by the instrument failure diagnose portion 42 from the process data collected by the data collection portion 41. The operation of the general fault diagnose portion 43 can be employed, for example, which is so arrange that .it -29outputs a diagnosis result as that this diagnoses result has a poor reliability upon the diagnosis item diagnosed by using the process value measured by the failed instrument by way of replacing the failed instrument's process value by an estimated value obtained by using the relationship upon the material balance or the heat balance. Another operation can be also employed which is so arranged that it outputs the diagnosis result as that the diagnosis result upon the diagnosis item cannot be obtained when the process value measured by the failed instrument is difficult to be estimated, but this process value performs a critical role in the diagnosis~ The output is, for example, in the form shown in Fig. 18.

According to the second embodiment, the process data collected by the data collection portion 41 is input to the instrument failure diagnose portion 42 and the general fault diagnose portion 43 so that the corresponding instrument failure and general fault are diagnosed. Even if an instrument failure is detected by the instrument Eailure diagnose portion 42, the general fault diagnose portion 43 performs the ault diaqnosis by removing the process data of the instrument diagnosed as failure from the process data collected by the data collection portion 41. Therefore, an operator can be ree from needless confusion.

Description will hereunder be given for a third embodiment according to the present invention. Before -3~explanation is made upon the third embodiment, a Eloating time and urgency factor according to this embodiment will be described. Even if faults evenly have the same time left to generation of fault state, their ~loating times are different to each other depending upon the time required taken to avoid occurrence of the fault state.

That is, a fault taking a rather long time period to avoid the fault state or restore a normal state is allowed a relatively short time since such fault needs to be early overcome. Therefore, the scale of such floating time can be expressed as follows:

floating time - (time left to generation of abnormal state) - ~time required to avoid abnormal state) ...(3~ Next, the urgency factor will now be described.

In the fault having short floating time, there is a fault having low importance factor and low urgency factor or a fault having high importance factor capable of critically affecting an apparatus and the surrounding and thereby having high urgency factor. There~ore, the scale of the urgency factor of the fault can be expressed as follows:

urgency factor = G (I, t) ... ~4) wherein I: importance factor when alarm is issued t : floating time for a certain alarm G : importance factor function of a certain alarm upon a floating time, and it satisfies the following characteristics:

G (I1, t) > G (I2, t), however I1 _ I2 G (I, t~ t2), however t1 _ t2 Referring to Fig. 19, reference numeral 51 repre~ents a data collection portion, reference numeral 52 represents a fault time estimating portion, reference numeral 53 represents an avoidance required time estimating po~tion. Reference numeral 54 represents a floating time estimating portion, and reference numeral 55 represents a floating time order display portion. A diagnose portion 60 is constituted by the fault time estimating portion 52, avoidance required time estimating portion 53, and a floating time estimating portion 54.

The data collection portion 51 acts to measure and collect data on the physical quantity o~ each portion of each process in the instrumentation of, for example, a chemical plant~ The fault time estimating portion 52 estimates the time taken of generation of a fault state upon an alarm item serving as a predetermined information ~or detecting fault or the like by using the physical properties of the process and the data collected by the ~7~ data collection portion 51. The thus-estimated value is called an "estimated fault-generation time". The avoidance required time estimating portion 53 estimates the time taken to avoid the subject fault state upon each of the predetermined alarm item in accordance with the physical properties of the process and data collected by the data collection portion 51. The thus-estimated value is called a "estimated avoidance-required time". The floating time estimating portion 54 calculates the floating time defined by the above-described equation (3). The floating time order display portion 55 displays the alarm item successively starting from the alarm item having short floating time in accordance with the floating time calculated by the floating time estimating portion 54.

The structure shown in Fig. 21 will then be described. The same portions as those shown in Fig. 19 are given the same reference numerals, and the descriptions for them are omitted. Re~erence numeral 56 represents an importance ~actor memory, reference numeral 57 represents an urgency ~actor estimating portion, and reference numeral 58 represents an urgency ~actor display portion serving as a noti~ying means. A diagnose portion 60 is constituted by the fault time estimating portion 52, the avoidance required tlme estimating portion 53, the floating time estimating portion 54, and the urgency factor estimating portion 57. In this arrangement, the importance factor memory 56 previously stores the importance factor when the alarm is issued upon alarm items serving as information for detecting fault or the like. The urgency factor estimatinq portion 57 calculates, by using equation (4~, the urgency factor by collating between the floating time calculated by the floating time estimating portion 54 and the importance factor when the alarm stored in the importance factor memory 56 is issued. The urgency factor display portion 58 displays successively starting from a fault state having the urgency factor in accordance with the urgency factor calculated by the urgency factor estimating portion 57.

By means of the above-described operation, the fault state which will be generated later having a rather short floating time or having a high urgency factor can be successively displayed. As a result of this, the instrumentation operator in, for example, a chemical plant, can previously acknowledge the fault state having short floating time or having a high emergency. It assists declsion making at the time of taking a countermeasure against a ~ault generated. As a result of khis, generation of faults in the instrumentation of the chemical plant due to the fault can be previously prevented.

The floating time order display portion 55 can further conveniently notify the fault state by addition 7~ ~34of factors to be displayed. Fig. 20 illustrates the structure of a system in this case. ThP same portion as those shown in Fig. 19 are given same reference numerals and their descriptions are omitted. In this structure, the floating time order display portion 55A is further provided with a time display function so that it displays the alarm factors starting from the alarm factor having a rather short floating time in accordance with the floating time calculated by the floating time estimating portion 54, and displays the estimated fault-generation time, the estimated avoidance-required time value and the estimated floating time depending upon each alarm factor.

As a result of thus-per~ormed display, because the absolute length of the floating time can be notified to humans and the estimated fault generation time and the estimated avoidance-required time are displayed, the progress of the fault in the instrumentation of the chemical plant can be easily notified to humans.

The urgency factor display portion 58 can be arranged to be further provided with factors to be displayed ~or the purpose of further convenient performlng notification of the fault state. Fig~ 22 illustrates the structure oP a system ~or use in this case. The same portions as those shown in Fig. 21 are given the same reference numerals, and the description for them are omitted. In this structure, a floating time and urgency factor display portion 59 is further provided -35with a time display function and importance factor display function so that it can successively display alarm factor starting from that having a large urgency factor in accordance with the urgency factor calculated by the urgency factor estimating portion 57, and it can display the estimated fault-generation time, estimated avoidance-required time, estimated floating time and urgency factor upon alarm factor. As a result of the thus-performed display, not only the sequential order of the urgency factor of the fault state but also progress and importance factor of the fault state can be further easily noticed to humans by displaying the estimated fault-generation time, estimated avoidance-required time, estimated floating time and importance factor.

A specific example applied to the instrumentation of a chemical plant according to the third embodiment will be further in detail described.

Fig. 23 illustrates a pump system serving as this specific example, in which a buffer chamber and a pump system are shown used in a case where fluid introduced into the system is branched into two systems.

In this figure, reference numeral 70 represents an inflow pipe, reference numerals 71a and 71b represent inflow indicator connected to the branched inflow pipes 70a and 70b. Reference numerals 72a and 72b represent inflow valve, and reference numerals 73a and 73b represent buffer chambers connected to the inflow pipes 70a and 7~3!35 -3~70b. Reference numerals 74a and 74b represent discharge pipes connected to the buffer chambers 73a and 73b, and reference numerals 75a and 75b represent level indicators for indicating the fluid level at the buffer cambers 73a and 73bo Reference numerals 76a and 76b represent discharge indicators connected to the discharge pipes 74a and 74b. Reference numerals 77a and 77b represent pumps connected to the discharge pipes 74a and 74b. These component elements constitutes a systems A and B.

The operation of the thus-constituted process will now be describedO Fluid is introduced from the inflow pipe 70 into the inflow pipes 70a and 70b, and is passed through the inflow indicators 71a and 71b and the inflow valves 72a and 72b. Then the fluid is introduced into the buffer chambers 73a and 73b. The inflow indicators 71a and 71b measure the rate of the flow passing into the buffer chambers 73a and 73b. The inflow valves 72a and 72b controls the rate of the flow introduced into the buffer chambers 73a and 73b. The buffer chambers 73a and 73b temporarily retain the fluid~ The fluid level in these buf~er chambers 73a and 73b are measured by the level indicators 75a and 75b. The fluid in the buffer chambers 73a and 73b is discharged through the discharge pipes 74a and 74b, and i5 passed through the discharge indicators 76a and 76b. The fluid is then delivered to separating devices 78a and 78b for separating the fluid into a light weight ingredient and a ` ~L;2'37~35 heavy weight ingredient by pumps 77a and 77b. The discharge indicators 76a and 7Çb mea~ure the flow rate passing through the discharge pipes 74a and 74b.

The fault diagnosing system according to the third embodiment is constituted for collecting dat~ of the measured values from the above-descrihed process by combining the inflow indicators 71a, 71b, inflow valves 72a and 72b, level indicators 75a and 76b and discharge indicators 76a and 76b.

The operation of each of the portions to which the system acaordlng to this embodiment in a case assumed that the inflow pipe 70 is clogged by foreign matters and the fluid rate passing from the inflow pipes 70 is thereby reduced will be described.

When the fluid rate supplied from the inflow pipe 70 is reduced, the quantity in the buffer chambers 73a and 73b is decreased assuming that the fluid quantity consumed in the lower stream portion of the subject process is not changed. Finally, the buffer chambers 73a and 73b become empty. As a result of this, the rate of the flow passing through the discharge pipes 74a and 74b is decreased, and Eluid supplied by the pumps 77a and 77b to the following stream becomes increased, causing the fault to be propagated to the following streams. The time taken for the buffer chambers 73a and 73b to become empty can be calculated from equation (5) using each residual quantity of the fluid in the buffer chambers 73a ~L2~9~ -38and 73b and the rate of the flow passing through the discharge pipes 74a and 74b. The systems A and B respectively involves the different time.

Ti = Qi (Li)/(Foi - Fni) ... (5) wherein Ti : time (second) taken for the buffer chamber in an i system becomes empty Qi : residual quantity (m3) of fluid in the buffer chamber in the i system, this quantity becoming a.

function o the residual fluid level Li in the buffer chamber Li : fluid level (%) in the buffer chamber in the i system Fni: rate (m3/second) of the flow passing though the i system i = a or b, each corresponding to the system A or B As a means for preventing an exhaustion of the buffer chambers 73a and 73b, the quantity reduced due to the clogglng of the 1n10w pipe 70 may be compensated by increase in the open degree of the inflow valves 72a and 72b ~or the purpose of reducing the pressure loss~ As the time required to avoid the fault, it i8 sufficiéntly safety to use the time taken to fully open the inflow valves 72a and 72b from the open degree of the --39-same.

The time required to fully open the valves can be obtained, for example, from equation (6~:

Tpi = Ai ~100 ~ Mi) ... (6) where in Tpi: time (second) taken to fully open the inflow valve in the system i Ai : constant ( second/ % ) determined by the characteristics of the valve Mi : .open degree ( % ) of the inf low valve in the system i . ..

i = a or b, each corresponding to the system A or B.

In this system, (A) is arranged to display an exhaustion of the buffer chamber 73a, while (B) is arranged to display the fault of the exhaustion of the buffer chambe}~ 73a.

The operation of each of the portions will now be described.

The data collection portion 51 collects the measured value of each instrument, that is, the rate of the f low passing through the inf low indicators 71 a and 71b, the fluid level of the level indicators 75a and 75b in the buf f er chambe.rs 7 3a and 7 3b, the rate of the f low passing through the discharge indicators 76a and 76b and the open degree of the inflow valves 72a and 72b.

The fault time estimatlng portion 52 calculates the time talcen for the f luid level in the buf fer chamber to become -40~ero, that is time Ti obtained from equation (5) so that this time is arranged to be the estimated value taken ~o generation of the fault.

The avoidance required time estimating portion 53 estimates the time required to take any avoidance measure, that isl it calculates Tpi given from equation (6) and makes this result as the estimated avoidance required time. The floating time estimating portion 54 subtracts the estimated avoidance re~uired time from the estimated fault time upon the faults in each of the systems and it makes this subtraction result the floating time. That is, abnormal A calculates Ta - Tpa upon the exhaustion of the buffer chamber 73a. Fault B calculates the exhaustion of the buffer chamber 73b. The floating time order display portion 55 arranges the fault factors having a short floating time to be at high ranking, and it, for example, displays the fault in accordance with this ranking on a cathode rays display to notify it to humans.

In the thus-constituted embodiment, either of buffer chambers in two systems earlier generates a fault can be previously acknowledged by monitoring them always so that a danger can be a~oided by flrst taking a certain measure for the system whose buffer chamber earlier becomes empty.

Fig. 24 illustrates another embodiment constituted by the system whose structures shown in Fig.

21 is combined.

Fig. 24 illustrates a buffer chamber and a pump systems with which the fluid introduced into these systems from inflow pipes are branched into two systems, and one of the two branches is directly supplied to the ensuing apparatus. The thus-formed system is arranged to be system Ao Another flow is not directly supplied to the ensuing device, but is stored in a chamber. This system is arranged to be system B. The same portions as those shown in Fig. 23 are given the same reference numerals, and the descriptions for them are omitted~ In this drawing, reference numeral 78 represents a device for fractionating the thus-branched fluid. Reference numeral 79 represents a chamber for storing the thusdivided 1uid. Since the operation of the thusconstituted process only has a difference in the portion to which the fluid is supplied from the process shown in Fig. 23, its description is omitted.

The operation of each of the portions to which the system according to this embodiment is subjected will be described assuming a case in which the fluid quantity supplied Erom the inflow pipe 70 is reduced due to clogging generated in the inElow pipe 70 by a foreign matter. As described above, the floating time for fault A and the floating time for fault B are calculated~ The importance factor when an aiarm is issued is previously set in the importance factor memory 56 in accordance with ~ 37~ -~2the classification criterion shown in Fig. 25. In this example, even if the fault (A) is generated, the apparatus cannot be broken. However, there is the probability of a large change in the ensuing fractionating apparatus or a deterioration in the product quality.

Therefore, its importance factor classification symbol corresponds to C. When the fault (B~ is generated, the apparatus cannot be broken and there is the little provability of deterioration in the product quality. However, since a slight change can be generated, its importance factor classification symbol corresponds to D. The urgency factor estimating portion 57 estimates the urgengy factor in accordance with the criterion shown in Fig. 26. In a case where the floating time for fault (A) is three minutes, while the floating time for fault (B) is two minutes, the urgency factor ~or the fault (A) is 4, while the same for the fault (B) is 5. Therefore, the faults ~A) having high urgency factor are successively dlsplayed on a display to notice it to humans.

As can be clearly seen from thls embodiment, since the ~ault havlng an urgency factor i8 clearly displayed considering the floating time of exhaustion fault the buffer chambers of the two systems and the effect when thls fault is generated by always monitoring the systems, it is clear for the operator to recognize the priority of the measures, therefore, the operator can 43properly takes a measure.

Next, a fourth embodiment of the present invention will now be described with reference to Fig.

27.

Referring to Fig. 27, this system comprises a sensor 81 such as a pressure gauge or a flow rate gauge for measuring the physical quantity of a specific point in the plant and for outputting its data. it further comprises an average value calculator 82A to which the output from the sensor 81 is input and calculating the average value of the physical quantity, and diagnose portion 83A to which the output from the average calculator 82A and the output from the sensor 81 are input, whereby the operating conditions of the plant is diagnosed by processing in a predetermined manner.

The average value calculator 82A is so arranged that the average value of the physical quantity in a time period which is immediately before a predetermined time period. The diagnosing portion 83A comprises a subtracter 86 for calculating the di~ference between the physical quantity at the specific point in the plant supplied from the sensor 81 and the average value supplied from the average value calculator 82A, and a comparator 87 for determining whether the difference obtained by th~ subtracter 86 is included in an allowable range which has been previously determined by an allowable range determinng device 85~ As a result of g~s -44this, the diagnosing portion 83A can perform fault diagnosis, and as well it comprises an alarm 88 serving as a notifying means for issuin~ an alarm when the difference between the physical quantity and the average value is not included in the allowable range. In this case, the average value calculated by the calculator 82A is considered to be the stable value of the physical quantity.

The average calculator 82A is arranged in such a manner that reset signals can be supplied if neaessary.

As a result of this, it can correspond to the change in the physical quantity generated when the operating conditions of the plant are changed.

Therefore, according to the fourth embodiment, since the allowable range can be set only for the deviation from the stable state, a relatively narrow allowable range can be set. Therefore, the problem experienced with the conventional apparatus in which the allowable range needs to be set wldely, that is, it takes a long time period for the change in the physlcal quantity due to the abnormality to exceed an allowable range can be overcome. Furthermore, since the average value serving as a re~erence value for a physical quantity in a stable state can correspond to the change in the process through a long time period, an effect can be obtained that lt can flexibly corresponds to the process change.

-~5 Furthermore, the average calculator 82A is constituted in such a manner that reset signals can be supplied and the average value of the physical quantity is calculated again from the time point at which such reset signal i~ supplied~ Therefore, even if the stable value of the physical quantity is changed due to the human factors such as change in the plant operating conditions, it can correspond to such change.

A modified examples of the fourth embodiment will now be described with reference to Figs. 28 to 31.

The same components parts as those in the fourth embodiment are given the same reference numerals, and the descriptions for them are omitted~ The modified example shown in Fig. 28 is characterized in that a hold circuit 89 is provided for the average calculator 82B. This hold circuit 89 acts not to renew the average value of the physical quantity when the alarm 88 issues an alarm in the immediately previous time range.

Therefore, according to the modified example, in addition to the efect obtained from the abovedescribed embodiment, a further eect can be obtained that a rather reliable alarm can be obtained since it becomes to use only the stable value o~ the physical quantity when the plant is normally operated can be used by arranging the system in which the average value of the physical quantity when any fault is detected in the plant \ :~Z~79~ -46is not used.

Fig. 29 illustrates the other modified example.

This modified example is chaxacterized in that the difference between the physical quantity measured by the sensor 81 and the average value of the physical quantity calculated by the average value calculator 82A, that is the range in which deviation distribu~es, is statistically calculatedO Therefore, an allowable range calculator 84A is as well provided between the average value calculator 82A and the comparator 87.

In this embodiment, assuming that the distribution of the above-described change is the normal distribution, the region in wh~ch the deviation distributes can be determined statistically from the following equation:

N m = ~ X (i) / N (7) i = 1 N ~ ( X ti)-m) 2 / t N - 1 ) (~) wherein X ~i). value measure of physical quantity at time point i W : the number of samples for the subjected physical quantity m : average value of the physical quantity : the standard deviation of X A : allowable range of the deviation ~LZ~317~ Assuming that the distribution of X is a normal distribution, the deviation is present in the probability o~ 99~ or more by making A = ~ 3a.

Therefore, by using this value for A as the allowable range, the fault o~ the physical quantity can be detected without any erroneous alarm is issued even when it is a normal or a stable state.

Furthermore, the allowable range for the deviation can be automatically determined and the allowable width deviated from the value at the normal operation can be automatically and significantly easily determined. Therefore, many labors required in the conventional system can be reduced to a considerable extent.

A modified example shown in Fig. 30 will now be described. The diagnosing portion 83B according to this modified example is characterlzed in that it is constituted in such a manner that an alarm is issued provided that the deviation between the physical quantity and the average value deviates ~or a predetermined continued time period.

Furthermore, a timer 90 is included~ This timer 90 is constituted in such a manner that it counts a timing when a signal representing generation of a devlation from an allowable range is input from the comparator, and outputs a signal demanding -48the operation of the alarm 88 to the alarm 88 when the deviation continues for a predetermined time period.

According to the thus-constituted example, an effect can be further obtained that an issue OI alarm can be prevented in a case where the change in the measured physical quantity is due to any disturbance such as an accidental noise or the like so that needless confusion can be prevented.

Fig. 31 illustrates the other modified example.

This modif ied example is characterized in that it is constituted in such a manner that reset signals can be intentionally supplied to the average value calculator 82C and allowable range calculator 84b so that the average value and the allowable range are calculated and renewed by using data which is the physical quantity ~rom receipt of these reset signals.

Therefore, according to the thus-constituted example, the average value oiE the physical quantity at the stable point intentionally changed and the devlation can be newly calculated, and the range in which the deviation is present corresponding to the operating conditions can be automatically determined. A8 a result o~ this, a ~ault detection having a wide applicability to various plants can be performed.

The plant fault diagnosing system according to the fourth embodiment and its modi ~ied examples are not limited to the case where the fault in the 37~ -49instrumentation in plants is directly detected and it is displayed~ Its diagnosis result can become the reference when the diagnosis pat-tern according to the first embodiment is processed. That is, when a fact that any fault is generated in the plant is diagnosed by the diagnose portions 83A and 83B, its diagnosis result is employed as the diagnosing pattern and as well the operating conditions or the like for the each portion of the plant are employed as the diagnosing pattern.

Although the sensor 81 comprises a pressure gauge and flow rate gauge, but it can be applied to various types of instrumentations widely used in plants~ As described above, according to the present invention, a fault diagnosing system can be provided with which faults and operating conditions of a plant can be early detected and monitored, and as well a reliable diagnosis result can be obtained. Furthermore, a countermeasures to always maintain the operating conditions of the plant can be taken by issuing an alarm or the like depending upon the diagnosis result.